Harden the elevation trust checks and narrow what one authorization applies

## Describe your changes

A group could be deleted while a reverse proxy service still referenced
it, silently breaking the service's access control: private services
list groups in `access_groups` as the peer allowlist, and SSO bearer
auth distributes tokens to `distribution_groups`.

Group deletion now runs through the same linkage validation as routes,
policies, and agent network policies: deleting a group that backs a
private service allowlist or an enabled bearer-auth distribution list
fails with a `GroupLinkError` naming the service domain. Disabled bearer
configs and stale `access_groups` on non-private services are inert and
do not block deletion.

Tests cover both linked cases in single and bulk deletion, and pin the
non-blocking cases. The test account seeds decoy services ahead of the
linked ones so the check is proven to scan the full service list.

.github/workflows/pr-title-check.yml

@@ -16,6 +16,8 @@ jobs:
             const allowedTags = [
               'management',
               'client',
+              'android',
+              'ios',
               'signal',
               'proxy',
               'relay',

.goreleaser_ui.yaml

@@ -92,8 +92,15 @@ nfpms:
         dst: /usr/share/applications/org.wails.netbird.desktop
       - src: client/ui/build/appicon.png
         dst: /usr/share/pixmaps/netbird.png
+      # Names the polkit action for the elevation prompt the app raises when an
+      # unprivileged user changes a privileged setting; without it the dialog
+      # shows a raw command line.
+      - src: client/ui/build/linux/polkit/io.netbird.settings.policy
+        dst: /usr/share/polkit-1/actions/io.netbird.settings.policy
     dependencies:
-      - netbird
+      - netbird (>= 0.75.0)
+      - libgtk-4-1 (>= 4.14)
+      - libwebkitgtk-6.0-4
 
   - maintainer: Netbird <dev@netbird.io>
     description: Netbird client UI.
@@ -113,8 +120,15 @@ nfpms:
         dst: /usr/share/applications/org.wails.netbird.desktop
       - src: client/ui/build/appicon.png
         dst: /usr/share/pixmaps/netbird.png
+      # Names the polkit action for the elevation prompt the app raises when an
+      # unprivileged user changes a privileged setting; without it the dialog
+      # shows a raw command line.
+      - src: client/ui/build/linux/polkit/io.netbird.settings.policy
+        dst: /usr/share/polkit-1/actions/io.netbird.settings.policy
     dependencies:
-      - netbird
+      - netbird >= 0.75.0
+      - (gtk4 >= 4.14 or libgtk-4-1 >= 4.14)
+      - (webkitgtk6.0 or libwebkitgtk-6_0-4)
 
     rpm:
       signature:

AGENTS.md

@@ -0,0 +1,514 @@
+# NetBird Agent Guidelines
+
+**NetBird** is an open-source connectivity platform: a WireGuard®-based overlay
+network with a control plane. The **agent** (`client/`) runs on user machines as
+a privileged daemon and manages the WireGuard interface, routing, firewall, and
+DNS. **Management** (`management/`) is the control plane and REST/gRPC API,
+**Signal** (`signal/`) brokers peer handshakes, **Relay** (`relay/`) carries
+traffic when a direct tunnel is impossible, and **Proxy** (`proxy/`) is the
+identity-aware proxy behind Agent Network.
+
+This file applies to the whole repository, and is the single source of truth for
+agent guidance here. `CLAUDE.md` is a one-line pointer to it — keep the guidance
+in this file, not duplicated there.
+
+## Contents
+
+- [NetBird Agent Guidelines](#netbird-agent-guidelines)
+  - [Contents](#contents)
+  - [STOP and ask the user before](#stop-and-ask-the-user-before)
+  - [Quick reference](#quick-reference)
+  - [Structure](#structure)
+  - [Where to look](#where-to-look)
+  - [Repo-wide principles](#repo-wide-principles)
+  - [Error handling](#error-handling)
+  - [Comments](#comments)
+  - [Testing](#testing)
+  - [Pitfalls](#pitfalls)
+  - [Commits, PRs, releases](#commits-prs-releases)
+  - [After you push: CI and review bots](#after-you-push-ci-and-review-bots)
+  - [Discussion and support](#discussion-and-support)
+
+## STOP and ask the user before
+
+- **Opening a pull request for anything beyond a trivial fix, without an agreed
+  ticket.** Ask the user directly: *"Is there a discussion or issue for this
+  change?"* NetBird is discussion-first — community reports start in
+  [Discussions](https://github.com/netbirdio/netbird/discussions), DevRel
+  validates them, and only validated discussions become issues. A PR that
+  changes behavior with no linked issue may be closed on arrival. If there is no
+  ticket, offer to draft the discussion post **instead of** the PR, and wait for
+  the user's call. Only typos, broken links, documentation corrections, and
+  one-line fixes that already have an issue can skip this.
+- **Designing in any high-risk area** (see
+  [CONTRIBUTING.md](CONTRIBUTING.md#high-risk-areas)): public API and OpenAPI
+  schema, gRPC protos, behavior existing deployments would notice after an
+  upgrade, peer connectivity (ICE, NAT traversal, relay selection, WireGuard® or
+  Rosenpass key handling), client system integration (routing, firewall, DNS,
+  interface), authentication and authorization, CLI or service flags, config
+  file format, daemon IPC, store schema and migrations, or a new feature. The
+  design gets agreed in the ticket before code is written.
+- **Writing a store migration or changing a persisted model.** Migrations are
+  one-way in the field and both the GORM and pgx paths may need the change.
+- **Hand-editing generated code.** `*.pb.go`, `*.gen.go`, and mocks are outputs.
+  Edit the source (`.proto`, `openapi.yml`) and rerun the matching
+  `generate.sh`.
+- **Adding, removing, or bumping a dependency**, and never vendor a fork.
+- **Weakening a security control** — authentication, authorization, certificate
+  verification, privilege dropping, or peer identity checks — even when it is
+  the fastest way to make a test pass.
+- **Force-pushing to `main`**, force-pushing any branch that is already under
+  review, amending pushed commits, or bypassing hooks with `--no-verify`.
+
+## Quick reference
+
+```bash
+# Build
+go build ./...
+cd client && CGO_ENABLED=0 go build .        # agent
+cd management && go build .                  # management service
+cd signal && go build .                      # signal service
+
+# Verify (run before every push)
+go fmt ./...
+make lint            # golangci-lint on files changed vs origin/main (also the pre-push hook)
+make lint-all        # full-repository lint, matches CI
+make test-unit       # host-safe unit tests, -tags devcert, no sudo
+make test-privileged # privileged-tagged suite in a Docker container with NET_ADMIN
+make setup-hooks     # wire make lint into .githooks/pre-push
+
+# Narrow runs
+go test ./client/internal/dns/...
+go test -race -run TestPeerConn ./client/internal/peer/...
+PRIV_RUN=TestNftablesManager PRIV_PKGS=./client/firewall/nftables/... make test-privileged
+
+# Code generation (never hand-edit the output)
+./shared/management/http/api/generate.sh   # REST types from openapi.yml
+./shared/management/proto/generate.sh
+./shared/signal/proto/generate.sh
+./client/proto/generate.sh
+./flow/proto/generate.sh
+
+# Run locally (lab only, never on a machine you rely on)
+sudo ./client/netbird up --log-level debug --log-file console
+sudo ./client/netbird down                   # teardown: restores routing, firewall, DNS
+./signal/signal run --log-level debug --log-file console
+./management/management management --log-level debug --log-file console --config ./management.json
+```
+
+`netbird up` needs root and rewrites the host's routing table, firewall rules,
+DNS configuration, and WireGuard® interface. Run it only in a disposable test
+environment (a VM, container, or throwaway host) that you can rebuild, never on
+a workstation or server whose connectivity matters. Run `sudo netbird down`
+before you stop working, before rebuilding the binary, and on every failure
+path, so the host's networking state is restored instead of left half-applied.
+See [Pitfalls](#pitfalls) for why cleanup on every exit path matters.
+
+## Structure
+
+```text
+netbird/
+├── client/              NetBird agent
+│   ├── cmd/             agent CLI
+│   ├── internal/        agent business logic (engine, peer, dns, routemanager, ...)
+│   ├── server/          daemon for background execution
+│   ├── proto/           daemon gRPC protos
+│   ├── iface/           WireGuard® interface management
+│   ├── firewall/        nftables, iptables, pf, WFP, userspace backends
+│   ├── ssh/             built-in SSH server and client
+│   ├── ui/              desktop UI (Wails v3 + React)
+│   ├── android/, ios/   mobile bindings
+│   ├── wasm/            WebAssembly build
+│   └── mdm/, system/    MDM policy, host information
+├── management/          control plane
+│   └── server/          account, peer, groups, networks, posture, permissions,
+│                        settings, store, http (REST), idp, integrations, migration
+├── signal/              handshake broker (peer/, server/)
+├── relay/               relay service (protocol/, server/, healthcheck/)
+├── proxy/               identity-aware proxy (llm/, acme/, accesslog/, middleware/, tcp/, udp/)
+├── agent-network/       Agent Network overview
+├── shared/              imported by both agent and services
+│   ├── management/      proto/, client/, http/api (OpenAPI + generated types)
+│   ├── signal/          proto/, client/
+│   └── relay/, auth/, sshauth/, metrics/
+├── e2e/                 end-to-end suites and harness
+├── encryption/, dns/, route/, stun/, sharedsock/, util/, flow/
+├── infrastructure_files/  docker compose and getting-started templates
+└── release_files/       files packaged into releases
+```
+
+## Where to look
+
+| Task                        | Location                                                     |
+| --------------------------- | ------------------------------------------------------------ |
+| REST API / OpenAPI          | `shared/management/http/api/` + `management/server/http/`    |
+| Management gRPC protocol    | `shared/management/proto/`                                   |
+| Signal protocol             | `shared/signal/proto/`                                       |
+| Daemon IPC protocol         | `client/proto/`                                              |
+| Peer connection and NAT     | `client/internal/peer/`                                      |
+| Network map handling        | `client/internal/engine.go`, `shared/management/networkmap/` |
+| Routing                     | `client/internal/routemanager/`, `route/`                    |
+| Firewall backends           | `client/firewall/`                                           |
+| DNS                         | `client/internal/dns/`, `dns/`                               |
+| WireGuard® interface        | `client/iface/`                                              |
+| Persistence and migrations  | `management/server/store/`, `management/server/migration/`   |
+| IdP integrations            | `management/server/idp/`                                     |
+| Permissions model           | `management/server/permissions/`                             |
+| LLM routing / Agent Network | `proxy/internal/llm/`, `agent-network/`                      |
+| End-to-end tests            | `e2e/`                                                       |
+
+## Repo-wide principles
+
+1. **Run `go fmt` on every modified Go file.** Formatting is not optional.
+2. **Zero unaddressed diagnostics.** Fix IDE and linter warnings on code you
+   touch, and delete imports, helpers, and parameters your refactor orphaned.
+   Exception: unused parameters in shared code may be consumed by builds outside
+   this repository — do not remove them, ask instead.
+3. **Function comments are mandatory for exported functions**, written as full
+   sentences with a period, starting with the identifier name.
+4. **Prefer private functions and constants.** Export only what a caller outside
+   the package genuinely needs.
+5. **Early returns and guard clauses.** Handle errors and edge cases first
+   instead of nesting `if`/`else` chains.
+6. **Split complex functions.** If a function trips a complexity warning, break
+   it into named helpers rather than silencing the warning.
+7. **Avoid LLM-slop tells:** em dashes, hedging narration, restating the diff in
+   prose, trailing summaries. Defaults, not absolute bans. Applies to code,
+   comments, commit messages, and PR descriptions alike.
+8. **Concurrency: do a two-pass race analysis after every change** that adds
+   shared state. Guard maps and slices with a mutex, keep critical sections
+   short, and run `go test -race` on the touched packages.
+9. **Cross-platform builds must keep working.** The agent targets Linux, macOS,
+   Windows, FreeBSD, Android, and iOS. When you add a platform-specific file,
+   add the counterpart or a build-tagged fallback for the others.
+10. **Never hand-edit generated files.** Change the source and regenerate.
+11. **Never log secrets** — private keys, setup keys, tokens, PAT values — and
+    keep peer IPs and hostnames out of logs above debug level.
+
+## Error handling
+
+Use single-assignment form when the error is only needed inside the `if`:
+
+```go
+// Good
+if err := someCall(); err != nil {
+    return fmt.Errorf("context: %w", err)
+}
+
+// Bad - unnecessary split
+err := someCall()
+if err != nil {
+    return fmt.Errorf("context: %w", err)
+}
+```
+
+Use multiple assignment when the value is needed after the block:
+
+```go
+result, err := someCall()
+if err != nil {
+    return fmt.Errorf("context: %w", err)
+}
+```
+
+Add short, meaningful context, and **do not** start `fmt.Errorf` messages with
+obvious words like "failed to" or "error":
+
+```go
+// Good
+return fmt.Errorf("parse remote address: %w", err)
+return fmt.Errorf("listen on %s: %w", addr, err)
+
+// Bad
+return fmt.Errorf("failed to parse remote address: %w", err)
+return fmt.Errorf("error listening on %s: %w", addr, err)
+
+// "failed" is fine in log messages
+log.Debugf("failed to parse remote address: %v", err)
+```
+
+Skip the wrapping when a function only extracts or delegates and the wrap would
+add nothing:
+
+```go
+func parseAddr(addr string) (string, int, error) {
+    host, portStr, err := net.SplitHostPort(addr)
+    if err != nil {
+        return "", 0, err
+    }
+    // ...
+}
+```
+
+Log the errors you choose not to act on:
+
+- `log.Debugf()` for errors that do not affect program flow but help debugging.
+- `log.Tracef()` for very verbose errors that would otherwise spam logs.
+- **Never ignore** errors from writes, network sends, or critical cleanup.
+- Close errors may be ignored for read-only operations; log them at debug for
+  writes.
+
+## Comments
+
+Comment the **why**, never the **what**. Default to no comment, and add one only
+when a hidden constraint or workaround would surprise a future reader. Never
+reference the current task, PR, or your own changes in a comment.
+
+```go
+// Bad - trailing comments explaining the obvious
+defer localConn.Close() // Close the connection
+if err != nil {         // Check if error occurred
+
+// Good
+defer localConn.Close()
+
+// Good - explains a non-obvious constraint
+// Use incremental checksum update per RFC 1624 for performance.
+checksum = updateChecksum(checksum, oldPort, newPort)
+```
+
+### Length budget
+
+- **90 characters per line.** Wrap the comment, do not run past it.
+- **250 characters per comment**, roughly three wrapped lines. Doc comments on
+  exported identifiers may exceed it when the API genuinely needs the
+  explanation; inline comments inside a function body may not.
+
+The budget is a smell detector, not a rule to game. Do not compress a needed
+explanation into cryptic shorthand to fit — if a block of code needs more than
+250 characters of prose, the code is doing too much. Fix the code:
+
+- **Extract a named function.** A well-named function replaces the comment: the
+  name says *what*, the body shows *how*, and the comment you no longer write
+  was the *what* anyway. Clean Code calls this "explain yourself in code".
+- **Extract a named constant or predicate.** `if isExpiredSetupKey(key)` needs
+  no comment; `if key.ExpiresAt.Before(now) && !key.Revoked && key.UsageLimit > 0`
+  does.
+- **Keep the surviving comment for the why** — the RFC, the kernel quirk, the
+  ordering constraint. That part is usually one or two lines.
+
+### Long switch and if/else chains
+
+A `switch` whose cases carry multi-line explanations is the usual place this
+budget is breached, and the comment is a symptom. In order of preference:
+
+1. **Extract each case body into a named function.** The case becomes one line,
+   the name carries the meaning, and the switch reads as a table of contents.
+2. **Replace the switch with a lookup table** — `map[Kind]handlerFunc` — when the
+   branches are uniform. Adding a case stops meaning editing a growing function.
+3. **Replace conditional with polymorphism** when branches vary by type and the
+   same switch shape starts appearing in more than one place. Clean Code's rule
+   of thumb: tolerate a switch statement if it appears **once**, is buried in a
+   factory that returns an interface, and no other switch dispatches on the same
+   type. A second switch over the same enum is the signal to introduce the
+   interface.
+
+Do not restructure a switch purely to satisfy the budget when the cases are one
+line each and self-evident — a flat, boring `switch` over an enum is fine and
+needs no comments at all.
+
+Explanatory comments in tests are welcome — they document the scenario being set
+up, and the 250-character budget does not apply to them.
+
+## Testing
+
+- **Unit tests** live beside the code as `_test.go`. `make test-unit` runs the
+  host-safe set with `-tags devcert` and no sudo.
+- **Privileged tests** carry the `privileged` build tag and mutate host
+  networking. They run through `make test-privileged`, inside a Docker container
+  with `NET_ADMIN`. Never bypass that harness by running them directly on the
+  host.
+- **End-to-end suites** live in `e2e/` with a shared harness.
+- **Test real behavior, not API existence.** Assert on the observable end state
+  a consumer would see — bytes that arrived, the packet after translation, the
+  row after the write — not merely that a method exists or returns an error.
+- **Avoid mocks for code we own.** Exercise the real store, manager, or
+  controller and assert what the caller actually receives.
+- **`require` for setup and preconditions, `assert` for the conditions under
+  test.** Use `require` whenever a later line would panic or be meaningless
+  otherwise.
+- **Message guidance:** optional for `NoError`/`Error`; always give context for
+  comparison, boolean, and collection assertions.
+
+```go
+server, err := StartTestServer()
+require.NoError(t, err, "Test server setup must succeed")
+defer server.Close()
+
+result, err := client.DoOperation()
+assert.NoError(t, err)
+assert.Equal(t, expectedResult, result, "Result should match expected")
+```
+
+## Pitfalls
+
+- **The agent runs as root.** Anything touching routing, firewall, DNS, or the
+  interface can take a user's machine off the network. Prefer a reversible
+  change and make sure cleanup runs on every exit path.
+- **Management has two account loaders** (GORM and pgx). Adding a relation to an
+  account often means updating both, or it silently comes back empty in
+  production.
+- **`go test ./...` without `-tags devcert` skips tests** that need the
+  development certificate. Use `make test-unit`.
+- **`make lint` only checks the diff against `origin/main`.** CI runs
+  `make lint-all`; run it too before pushing a large change.
+- **Protos are consumed by released clients.** An old agent must keep working
+  against a new Management, so fields are added, never renumbered or removed.
+- **Windows requires the wintun driver**, and the daemon serves a named pipe
+  (`npipe://netbird`) rather than loopback TCP. Loopback TCP carries no caller
+  identity, so privileged operations are refused over it.
+
+## Commits, PRs, releases
+
+- **PR titles must start with a bracketed tag.** Before you propose a title,
+  **read [`.github/workflows/pr-title-check.yml`](.github/workflows/pr-title-check.yml)
+  and take the allowed tags from the `allowedTags` array in that file.** It is
+  the only source of truth, it changes as components are added, and the check
+  runs on every title edit — a tag that is not in that array is a red build. Do
+  not rely on a list memorized from anywhere else, including this file.
+
+  ```text
+  [client] Authorize daemon IPC callers by their local identity
+  [management,client] Add MDM policy support
+  ```
+
+  Multiple tags are comma-separated inside one pair of brackets. Match the tag
+  to the component you actually changed, not to the one you read the most.
+
+- **Use the repository's PR template.** Fill in
+  [`.github/pull_request_template.md`](.github/pull_request_template.md) rather
+  than replacing it with your own summary: describe the change, link the issue,
+  tick the checklist honestly (including "ran locally" and "single purpose"),
+  and complete the documentation section. Do not tick a box you have not
+  verified, and do not delete rows that do not apply.
+
+- **Keep the PR description short.** Under 1000 words on top of the template's
+  own text, and usually far less — a few paragraphs. Reviewers read the diff;
+  the description exists to explain what the diff cannot say for itself. This is
+  well below what an agent will produce by default, so cut before you post.
+
+- **Body: why before what.** Lead with the problem and the reason for this
+  approach, then the shape of the change. No bullet list of files changed, no
+  per-function walkthrough, no restating the diff in prose, no trailing summary
+  section, no self-congratulatory closing line.
+
+- **No `Co-Authored-By` or tool-attribution trailers in the PR description**,
+  and none in commits either. Contributors own their contributions. Whatever
+  tooling produced the diff, the person opening the PR is its author: they have
+  read every line, they can explain why it works, they can answer review
+  questions without going back to a model, and they are accountable for the
+  consequences of merging it. Do not add a trailer, footer, or description line
+  that spreads that ownership onto a tool.
+
+- **Commit subjects follow the same `[scope] Subject` convention.** Keep the
+  subject short, and use the body for why before what. No bullet lists of files
+  changed.
+
+- **Push review fixes as separate commits.** The PR is squashed on merge, so
+  there is no reason to rewrite history mid-review; many small commits make the
+  re-review readable.
+
+- **Do not force-push a branch that is under review.** A force-push detaches
+  existing review comments from the lines they were written against, destroys
+  the "changes since your last review" diff a reviewer relies on, and discards
+  the CI history that showed which commit broke what. Add commits instead —
+  including for fixups and reverts. Force-push only when there is no
+  alternative: a rebase to clear a genuine conflict, or removing a secret or a
+  large binary that was committed by mistake. When you must, ask the user first,
+  then say so in a PR comment so reviewers know their anchors moved. Never
+  force-push `main`, and never force-push a branch you do not own.
+
+- **One PR, one purpose.** Split refactors out of fixes and fixes out of
+  features.
+
+- **Keep the PR small.** Size is the single strongest predictor of how long a PR
+  waits. Aim for **under ~400 changed lines across under ~20 files**; past
+  roughly **1000 lines or 50 files** a community PR is likely to be sent back to
+  be split, or left unreviewed until it is. Large PRs from outside the core team
+  may be blocked outright when the size was never agreed in the ticket —
+  reviewing a sprawling change against a privileged networking daemon is a
+  security risk in itself, not just a time cost.
+
+  Judge the size by hand-written code: exclude generated output, `go.sum`,
+  vendored files, and test fixtures from the estimate, but do not use their
+  presence to argue a 3000-line PR is small.
+
+  When a change genuinely cannot be small — a protocol migration, a
+  cross-component rename — agree the split in the ticket **before** writing
+  code, and land it as a sequence of PRs that each build, test, and make sense
+  on their own. Propose that split to the user rather than opening one large PR
+  and hoping.
+
+- **User-facing changes need a docs PR** in
+  [netbirdio/docs](https://github.com/netbirdio/docs), linked from the PR
+  description.
+
+## After you push: CI and review bots
+
+Opening the PR is not the end of the task. Watch the run, read what the bots
+say, and drive the PR to green before you report the work as done.
+
+```bash
+gh pr checks <pr>            --watch    # all checks, live
+gh run view <run-id> --log-failed       # only the failing steps
+gh pr view <pr> --comments              # bot and human review comments
+```
+
+**Never report a change as finished while checks are pending or red**, and never
+describe a red PR as passing. If you ran out of turn before CI finished, say
+which checks were still running.
+
+### The checks
+
+- **Go tests** — `golang-test-{linux,darwin,windows,freebsd}.yml`, sharded per
+  component. A failure in a component you did not touch is usually a real
+  interaction, not noise; read the log before assuming flake.
+- **golangci-lint** — `golangci-lint.yml` runs the full repository, while
+  `make lint` only checks your diff. A clean local lint does not guarantee green
+  CI on a large change.
+- **PR Title Check** — `pr-title-check.yml`, see above.
+- **Codecov** — uploaded from the Linux test workflow with per-component flags
+  (`unit,client`, `unit,management`, `unit,relay`, `unit,proxy`, `unit,signal`,
+  `integration,management`). Coverage on new code should not go backwards. Add
+  tests for the paths you introduced; do not adjust thresholds or exclude files
+  to clear the report.
+- **CodeRabbit** — configured in [`.coderabbit.yaml`](.coderabbit.yaml): `chill`
+  profile, auto-review on every non-draft PR, TypeScript/JavaScript/SVG paths
+  filtered out. Chat auto-reply is on, so `@coderabbitai` in a comment reaches
+  it.
+- **SonarCloud** — project `netbirdio_netbird`, quality gate on new code (bugs,
+  vulnerabilities, code smells, duplication, coverage).
+- **Snyk** — dependency and code scanning.
+
+Sonar and Snyk report as GitHub App checks rather than workflows in this
+repository, so their detail lives on the PR check, not in the Actions logs.
+
+### Handling bot findings
+
+- **Read every comment and act on it.** Either fix it, or reply with the reason
+  it does not apply. Do not bulk-resolve threads to clear the count, and do not
+  silently ignore a finding because the check is advisory.
+- **Bots are frequently wrong here.** NetBird has privileged, platform-specific,
+  and concurrency-heavy code that static analysis reads poorly. A confident
+  CodeRabbit or Sonar comment can still be nonsense. Verify the claim against
+  the code before you change anything — never edit correct code just to silence
+  a bot.
+- **Security findings get the opposite default.** For a Snyk or Sonar
+  vulnerability, or a CodeRabbit comment about authentication, authorization,
+  certificate verification, or key handling, assume it is real until you have
+  disproved it. Surface it to the user rather than dismissing it yourself.
+- **A new vulnerable dependency is a stop.** Bumping or replacing dependencies
+  needs the user's decision, as above.
+- **Never change a workflow, threshold, lint exclusion, or bot config to make a
+  check pass.** If a check is genuinely wrong, say so and let the user decide.
+- **Do not paper over flakes with blind re-runs.** Identify the failure first. If
+  it is a known flake, name it; if you cannot tell, report it as unresolved
+  rather than re-running until it goes green.
+
+## Discussion and support
+
+- Discussions: <https://github.com/netbirdio/netbird/discussions>
+- Slack: <https://docs.netbird.io/slack-url>
+- Docs: <https://docs.netbird.io>
+- Security: <https://github.com/netbirdio/netbird/security/policy> — never in public
+- Contribution process: [CONTRIBUTING.md](CONTRIBUTING.md)

CLAUDE.md

@@ -0,0 +1 @@
+See [AGENTS.md](AGENTS.md) for the agent guidelines in this repository.

CONTRIBUTING.md

@@ -66,11 +66,41 @@ Typical bug fixes, internal refactors, documentation updates, and tests do not
 need a design discussion, but should still be tied to an issue so the work is
 visible and nobody duplicates it.
 
+### Using AI coding agents
+
+We have no policy for or against using an AI agent to write NetBird code. That
+choice is yours, and we are not going to interrogate anyone about their tools.
+
+What we do have is a lot of incoming contributions that were plainly drafted with
+one, and enough experience reviewing them to see the same avoidable problems
+again and again: no ticket behind the change, a diff far too large to review, a
+description longer than the code it describes, an approach that was never going
+to be accepted, and an author who cannot answer questions about their own PR.
+None of that is caused by the tooling — it is what happens when a tool is pointed
+at a repository whose expectations it has never been told.
+
+So rather than a rule, there is a guide. [AGENTS.md](AGENTS.md) restates the
+expectations from this document in the form agents read automatically
+(`CLAUDE.md` points to it), so pointing your tool at the repository is usually
+enough. Among other things it tells the agent to ask you for the
+discussion or issue before drafting a PR, to keep the change small and
+single-purpose, to run the tests locally, to use this repository's PR template
+and title tags, and to write a description a reviewer can get through.
+
+The guardrails are the point, and they are the same ones we apply to everyone: an
+agreed ticket, a change you have actually run, a diff small enough to review with
+care, and an author who can explain it. Whatever wrote the diff, you are its
+author — you own every line you submit and the consequences of opening a PR with it.
+
+We may assess whether a contribution is maintainable and whether its merged code
+aligns with our security standards and design expectations.
+
 ## Contents
 
 - [Contributing to NetBird](#contributing-to-netbird)
     - [Ticket first, PR second](#ticket-first-pr-second)
         - [High-risk areas](#high-risk-areas)
+        - [Using AI coding agents](#using-ai-coding-agents)
     - [Contents](#contents)
     - [Code of conduct](#code-of-conduct)
     - [Directory structure](#directory-structure)

client/android/client.go

@@ -57,6 +57,12 @@ type DnsReadyListener interface {
 	dns.ReadyListener
 }
 
+// TunSettings is a snapshot of the settings the TUN device is rebuilt with
+type TunSettings struct {
+	Routes        string
+	SearchDomains string
+}
+
 func init() {
 	formatter.SetLogcatFormatter(log.StandardLogger())
 }
@@ -76,6 +82,8 @@ type Client struct {
 	connectClient *internal.ConnectClient
 	config        *profilemanager.Config
 	cacheDir      string
+	// Identifies the running profile for the SSO login hint; see profile_state.go.
+	cfgPath string
 
 	stateChangeMu    sync.Mutex
 	stateChangeSubID string
@@ -96,11 +104,12 @@ type Client struct {
 	extendCancel context.CancelFunc
 }
 
-func (c *Client) setState(cfg *profilemanager.Config, cacheDir string, cc *internal.ConnectClient) {
+func (c *Client) setState(cfg *profilemanager.Config, cacheDir string, cfgPath string, cc *internal.ConnectClient) {
 	c.stateMu.Lock()
 	defer c.stateMu.Unlock()
 	c.config = cfg
 	c.cacheDir = cacheDir
+	c.cfgPath = cfgPath
 	c.connectClient = cc
 }
 
@@ -110,6 +119,16 @@ func (c *Client) stateSnapshot() (*profilemanager.Config, string, *internal.Conn
 	return c.config, c.cacheDir, c.connectClient
 }
 
+// authSnapshot returns the config together with the path it was loaded from, in
+// one lock: the path identifies the profile whose account email backs the login
+// hint, so reading it separately could pair one profile's config with another's
+// hint when a profile switch lands in between.
+func (c *Client) authSnapshot() (*profilemanager.Config, string, *internal.ConnectClient) {
+	c.stateMu.RLock()
+	defer c.stateMu.RUnlock()
+	return c.config, c.cfgPath, c.connectClient
+}
+
 func (c *Client) getConnectClient() *internal.ConnectClient {
 	c.stateMu.RLock()
 	defer c.stateMu.RUnlock()
@@ -162,7 +181,7 @@ func (c *Client) Run(platformFiles PlatformFiles, urlOpener URLOpener, isAndroid
 	defer c.ctxCancel()
 	c.ctxCancelLock.Unlock()
 
-	auth := NewAuthWithConfig(ctx, cfg)
+	auth := NewAuthWithConfig(ctx, cfg, cfgFile)
 	err = auth.login(urlOpener, isAndroidTV)
 	if err != nil {
 		return err
@@ -170,7 +189,7 @@ func (c *Client) Run(platformFiles PlatformFiles, urlOpener URLOpener, isAndroid
 	// todo do not throw error in case of cancelled context
 	ctx = internal.CtxInitState(ctx)
 	connectClient := internal.NewConnectClient(ctx, cfg, c.recorder)
-	c.setState(cfg, cacheDir, connectClient)
+	c.setState(cfg, cacheDir, cfgFile, connectClient)
 	// This path runs the interactive SSO flow, so reaching here means the peer
 	// is authenticated again — release the latch Status() reports from. Clear
 	// only once the fresh connect client is installed: until then Status()
@@ -211,7 +230,7 @@ func (c *Client) RunWithoutLogin(platformFiles PlatformFiles, dns *DNSList, dnsR
 	// todo do not throw error in case of cancelled context
 	ctx = internal.CtxInitState(ctx)
 	connectClient := internal.NewConnectClient(ctx, cfg, c.recorder)
-	c.setState(cfg, cacheDir, connectClient)
+	c.setState(cfg, cacheDir, cfgFile, connectClient)
 	return connectClient.RunOnAndroid(c.tunAdapter, c.iFaceDiscover, c.networkChangeListener, slices.Clone(dns.items), dnsReadyListener, stateFile, cacheDir)
 }
 
@@ -240,6 +259,24 @@ func (c *Client) RenewTun(fd int) error {
 	return e.RenewTun(fd)
 }
 
+func (c *Client) GetTunSettings() (*TunSettings, error) {
+	cc := c.getConnectClient()
+	if cc == nil {
+		return nil, fmt.Errorf("engine not running")
+	}
+
+	e := cc.Engine()
+	if e == nil {
+		return nil, fmt.Errorf("engine not initialized")
+	}
+
+	routes, searchDomains := e.TunSettings()
+	return &TunSettings{
+		Routes:        strings.Join(routes, ";"),
+		SearchDomains: strings.Join(searchDomains, ";"),
+	}, nil
+}
+
 // DebugBundle generates a debug bundle, uploads it, and returns the upload key.
 // It works both with and without a running engine.
 func (c *Client) DebugBundle(platformFiles PlatformFiles, anonymize bool) (string, error) {

client/android/login.go

@@ -4,6 +4,8 @@ import (
 	"context"
 	"fmt"
 
+	log "github.com/sirupsen/logrus"
+
 	"github.com/netbirdio/netbird/client/internal/auth"
 	"github.com/netbirdio/netbird/client/internal/profilemanager"
 	"github.com/netbirdio/netbird/client/system"
@@ -36,12 +38,20 @@ type Auth struct {
 }
 
 // NewAuth instantiate Auth struct and validate the management URL
+//
+// The configuration at cfgPath is reused when one is already there, and only created when it is
+// not. Building a fresh in-memory config unconditionally gives the client a new WireGuard key on
+// every call: the peer registers under that key, the key is written out, and any peer registered by
+// an earlier call is orphaned on the server. It also breaks a client that enrols and then runs from
+// the persisted config, because the identity it registered is not the one it runs with — the
+// management stream rejects it with "no peer auth method provided".
 func NewAuth(cfgPath string, mgmURL string) (*Auth, error) {
 	inputCfg := profilemanager.ConfigInput{
+		ConfigPath:    cfgPath,
 		ManagementURL: mgmURL,
 	}
 
-	cfg, err := profilemanager.CreateInMemoryConfig(inputCfg)
+	cfg, err := profilemanager.UpdateOrCreateConfig(inputCfg)
 	if err != nil {
 		return nil, err
 	}
@@ -53,11 +63,14 @@ func NewAuth(cfgPath string, mgmURL string) (*Auth, error) {
 	}, nil
 }
 
-// NewAuthWithConfig instantiate Auth based on existing config
-func NewAuthWithConfig(ctx context.Context, config *profilemanager.Config) *Auth {
+// NewAuthWithConfig instantiate Auth based on existing config. cfgPath is the
+// file the config was loaded from; it identifies the profile whose account email
+// backs the login_hint.
+func NewAuthWithConfig(ctx context.Context, config *profilemanager.Config, cfgPath string) *Auth {
 	return &Auth{
-		ctx:    ctx,
-		config: config,
+		ctx:     ctx,
+		config:  config,
+		cfgPath: cfgPath,
 	}
 }
 
@@ -150,12 +163,14 @@ func (a *Auth) login(urlOpener URLOpener, isAndroidTV bool) error {
 	}
 
 	jwtToken := ""
+	email := ""
 	if needsLogin {
 		tokenInfo, err := a.foregroundGetTokenInfo(authClient, urlOpener, isAndroidTV)
 		if err != nil {
 			return fmt.Errorf("interactive sso login failed: %v", err)
 		}
 		jwtToken = tokenInfo.GetTokenToUse()
+		email = tokenInfo.Email
 	}
 
 	err, _ = authClient.Login(a.ctx, "", jwtToken)
@@ -163,17 +178,42 @@ func (a *Auth) login(urlOpener URLOpener, isAndroidTV bool) error {
 		return fmt.Errorf("login failed: %v", err)
 	}
 
+	// Stored after Login, not before: a rejected token must not leave a hint
+	// pointing at an account that cannot be used.
+	if email != "" && a.cfgPath != "" {
+		if err := writeProfileEmail(a.cfgPath, email); err != nil {
+			log.Warnf("failed to store profile account email: %v", err)
+		}
+	}
+
 	go urlOpener.OnLoginSuccess()
 
 	return nil
 }
 
+// loginHintSetter is implemented by both concrete flows (PKCE and device code)
+// but absent from the OAuthFlow interface, hence the assertion below — the same
+// way internal/auth wires it in authenticateWithPKCEFlow.
+type loginHintSetter interface {
+	SetLoginHint(hint string)
+}
+
 func (a *Auth) foregroundGetTokenInfo(authClient *auth.Auth, urlOpener URLOpener, isAndroidTV bool) (*auth.TokenInfo, error) {
 	oAuthFlow, err := authClient.GetOAuthFlow(a.ctx, isAndroidTV)
 	if err != nil {
 		return nil, fmt.Errorf("failed to get OAuth flow: %v", err)
 	}
 
+	// An empty hint is deliberate, not a fallback: a fresh or logged-out profile
+	// leaves the choice to the IdP, which is how accounts get switched.
+	if a.cfgPath != "" {
+		if hint := readProfileEmail(a.cfgPath); hint != "" {
+			if setter, ok := oAuthFlow.(loginHintSetter); ok {
+				setter.SetLoginHint(hint)
+			}
+		}
+	}
+
 	flowInfo, err := oAuthFlow.RequestAuthInfo(context.TODO())
 	if err != nil {
 		return nil, fmt.Errorf("getting a request OAuth flow info failed: %v", err)

client/android/login_test.go

@@ -0,0 +1,51 @@
+package android
+
+import (
+	"path/filepath"
+	"testing"
+)
+
+// NewAuth must reuse the configuration already at cfgPath rather than building a fresh one.
+//
+// Creating a new in-memory config on every call gives the client a new WireGuard private key each
+// time. The peer registers under that key and the key is written out, so a peer registered by an
+// earlier call is orphaned on the server — a client that enrols twice leaves two entries and owns
+// neither. It also breaks enrol-then-run: RunWithoutLogin reloads the configuration from disk, so
+// the identity that registered is not the identity that runs, and the management stream rejects it
+// with "no peer auth method provided, please use a setup key or interactive SSO login".
+func TestNewAuth_ReusesPersistedIdentity(t *testing.T) {
+	cfgPath := filepath.Join(t.TempDir(), "config.json")
+
+	first, err := NewAuth(cfgPath, "https://api.example.com:443")
+	if err != nil {
+		t.Fatalf("first NewAuth: %v", err)
+	}
+	if first.config.PrivateKey == "" {
+		t.Fatal("first NewAuth produced no private key")
+	}
+
+	second, err := NewAuth(cfgPath, "https://api.example.com:443")
+	if err != nil {
+		t.Fatalf("second NewAuth: %v", err)
+	}
+
+	if second.config.PrivateKey != first.config.PrivateKey {
+		t.Errorf("private key changed between calls: a second enrolment would orphan the peer registered by the first")
+	}
+}
+
+// A missing configuration is still created, so a first enrolment works unchanged.
+func TestNewAuth_CreatesConfigWhenAbsent(t *testing.T) {
+	cfgPath := filepath.Join(t.TempDir(), "config.json")
+
+	auth, err := NewAuth(cfgPath, "https://api.example.com:443")
+	if err != nil {
+		t.Fatalf("NewAuth: %v", err)
+	}
+	if auth.config == nil || auth.config.PrivateKey == "" {
+		t.Fatal("NewAuth did not create a usable configuration")
+	}
+	if auth.cfgPath != cfgPath {
+		t.Errorf("cfgPath = %q, want %q", auth.cfgPath, cfgPath)
+	}
+}

client/android/profile_manager.go

@@ -13,18 +13,17 @@ import (
 )
 
 const (
-	// Android-specific config filename (different from desktop default.json)
-	defaultConfigFilename = "netbird.cfg"
-	// Subdirectory for non-default profiles (must match Java Preferences.java)
-	profilesSubdir = "profiles"
 	// Android uses a single user context per app (non-empty username required by ServiceManager)
 	androidUsername = "android"
 )
 
 // Profile represents a profile for gomobile
 type Profile struct {
-	ID       string
-	Name     string
+	ID   string
+	Name string
+	// Email is the account this profile last logged in with, "" if it never
+	// completed an SSO login or was logged out. See profile_state.go.
+	Email    string
 	IsActive bool
 }
 
@@ -101,6 +100,7 @@ func (pm *ProfileManager) ListProfiles() (*ProfileArray, error) {
 		profiles = append(profiles, &Profile{
 			ID:       p.ID.String(),
 			Name:     p.Name,
+			Email:    pm.profileEmail(p.ID.String()),
 			IsActive: p.IsActive,
 		})
 	}
@@ -123,7 +123,22 @@ func (pm *ProfileManager) GetActiveProfile() (*Profile, error) {
 	if err != nil {
 		return nil, fmt.Errorf("failed to resolve active profile %q: %w", activeState.ID, err)
 	}
-	return &Profile{ID: prof.ID.String(), Name: prof.Name, IsActive: true}, nil
+	return &Profile{
+		ID:       prof.ID.String(),
+		Name:     prof.Name,
+		Email:    pm.profileEmail(prof.ID.String()),
+		IsActive: true,
+	}, nil
+}
+
+// profileEmail returns the account email recorded for a profile. Display-only, so
+// an unresolvable path degrades to "" rather than an error.
+func (pm *ProfileManager) profileEmail(id string) string {
+	configPath, err := pm.getProfileConfigPath(id)
+	if err != nil {
+		return ""
+	}
+	return readProfileEmail(configPath)
 }
 
 // SwitchProfile switches to a different profile
@@ -185,6 +200,11 @@ func (pm *ProfileManager) LogoutProfile(id string) error {
 		return fmt.Errorf("failed to save config: %w", err)
 	}
 
+	// Not fatal: a stale hint costs an account switch, not the logout itself.
+	if err := removeProfileEmail(configPath); err != nil {
+		log.Warnf("failed to clear stored account email for profile %s: %v", id, err)
+	}
+
 	log.Infof("logged out from profile: %s", id)
 	return nil
 }

client/android/profile_state.go

@@ -0,0 +1,108 @@
+package android
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+
+	log "github.com/sirupsen/logrus"
+
+	"github.com/netbirdio/netbird/client/internal/profilemanager"
+	"github.com/netbirdio/netbird/util"
+)
+
+const (
+	// Android-specific config filename (different from desktop default.json)
+	defaultConfigFilename = "netbird.cfg"
+	// Subdirectory for non-default profiles (must match Java Preferences.java)
+	profilesSubdir = "profiles"
+	// profileAccountSuffix names the file holding the profile's account email.
+	// Deliberately not ".state.json", which desktop uses for the same data:
+	// there the email and the engine's state manager live in different
+	// directories, but on Android both resolve under files/, so sharing the name
+	// would have the two overwrite each other — the state manager rewrites the
+	// whole file from its own keys (see statemanager.Manager.PersistState), and
+	// this package's writer does the same in reverse.
+	profileAccountSuffix = ".account.json"
+)
+
+// profileAccountPathFor derives the account file path from a profile's config
+// path: netbird.cfg -> netbird.account.json, <id>.json -> <id>.account.json.
+//
+// Deriving from the config path rather than resolving the active profile keeps
+// the write on the profile the login actually ran for: Auth.login runs in a
+// goroutine, so the active profile can change under a flow already in flight.
+func profileAccountPathFor(configPath string) (string, error) {
+	if configPath == "" {
+		return "", fmt.Errorf("empty config path")
+	}
+
+	base := filepath.Base(configPath)
+	stem := strings.TrimSuffix(base, filepath.Ext(base))
+	if stem == "" || stem == "." {
+		return "", fmt.Errorf("config path %q has no filename stem", configPath)
+	}
+
+	return filepath.Join(filepath.Dir(configPath), stem+profileAccountSuffix), nil
+}
+
+// readProfileEmail returns the account email stored for the profile whose config
+// lives at configPath. A missing or unreadable file yields "", which leaves the
+// account choice to the IdP.
+func readProfileEmail(configPath string) string {
+	accountPath, err := profileAccountPathFor(configPath)
+	if err != nil {
+		log.Debugf("no profile account path for login hint: %v", err)
+		return ""
+	}
+
+	var state profilemanager.ProfileState
+	if _, err := util.ReadJson(accountPath, &state); err != nil {
+		if !os.IsNotExist(err) {
+			log.Debugf("failed to read profile account for login hint: %v", err)
+		}
+		return ""
+	}
+
+	return state.Email
+}
+
+// writeProfileEmail records the account email for the profile whose config lives
+// at configPath, so later logins can pass it as an OIDC login_hint. An empty
+// email is ignored rather than blanking what is already stored.
+func writeProfileEmail(configPath string, email string) error {
+	if email == "" {
+		return nil
+	}
+
+	accountPath, err := profileAccountPathFor(configPath)
+	if err != nil {
+		return fmt.Errorf("resolve profile account path: %w", err)
+	}
+
+	state := profilemanager.ProfileState{Email: email}
+	if err := util.WriteJsonWithRestrictedPermission(context.Background(), accountPath, state); err != nil {
+		return fmt.Errorf("write profile account: %w", err)
+	}
+
+	return nil
+}
+
+// removeProfileEmail drops the stored account email. Called on logout: while the
+// email is on disk it goes out as a login_hint, which would steer the next login
+// straight back into the account just logged out of. Mirrors the desktop UI's
+// RemoveProfileState call.
+func removeProfileEmail(configPath string) error {
+	accountPath, err := profileAccountPathFor(configPath)
+	if err != nil {
+		return fmt.Errorf("resolve profile account path: %w", err)
+	}
+
+	if err := os.Remove(accountPath); err != nil && !os.IsNotExist(err) {
+		return fmt.Errorf("remove profile account: %w", err)
+	}
+
+	return nil
+}

client/android/profile_state_test.go

@@ -0,0 +1,161 @@
+package android
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+)
+
+func TestProfileAccountPathFor(t *testing.T) {
+	tests := []struct {
+		name       string
+		configPath string
+		want       string
+		wantErr    bool
+	}{
+		{
+			name:       "default profile",
+			configPath: "/data/data/io.netbird.client/files/netbird.cfg",
+			want:       filepath.FromSlash("/data/data/io.netbird.client/files/netbird.account.json"),
+		},
+		{
+			name:       "id profile",
+			configPath: "/data/data/io.netbird.client/files/profiles/4c5f5c8198c3989cffb5b5394f5a7ae0.json",
+			want:       filepath.FromSlash("/data/data/io.netbird.client/files/profiles/4c5f5c8198c3989cffb5b5394f5a7ae0.account.json"),
+		},
+		{
+			name:       "legacy name-keyed profile is handled the same way",
+			configPath: "/data/data/io.netbird.client/files/profiles/work.json",
+			want:       filepath.FromSlash("/data/data/io.netbird.client/files/profiles/work.account.json"),
+		},
+		{
+			name:       "empty path is rejected",
+			configPath: "",
+			wantErr:    true,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			got, err := profileAccountPathFor(tt.configPath)
+			if tt.wantErr {
+				if err == nil {
+					t.Fatalf("expected an error, got path %q", got)
+				}
+				return
+			}
+			if err != nil {
+				t.Fatalf("unexpected error: %v", err)
+			}
+			if got != tt.want {
+				t.Errorf("got %q, want %q", got, tt.want)
+			}
+		})
+	}
+}
+
+func TestProfileAccountPathForDefaultDoesNotCollide(t *testing.T) {
+	root := "/data/data/io.netbird.client/files"
+
+	defaultAccount, err := profileAccountPathFor(filepath.Join(root, defaultConfigFilename))
+	if err != nil {
+		t.Fatalf("default profile: %v", err)
+	}
+
+	idAccount, err := profileAccountPathFor(filepath.Join(root, profilesSubdir, "abc123.json"))
+	if err != nil {
+		t.Fatalf("id profile: %v", err)
+	}
+
+	if defaultAccount == idAccount {
+		t.Fatalf("default and id profile share an account file: %q", defaultAccount)
+	}
+}
+
+// The account file must never land on the engine state file: on Android both
+// resolve under files/, and the state manager rewrites the whole file from its
+// own keys, so sharing a path would have the two overwrite each other. The
+// expected names here mirror ProfileManager.GetStateFilePath.
+func TestProfileAccountPathAvoidsEngineStateFile(t *testing.T) {
+	root := "/data/data/io.netbird.client/files"
+
+	cases := []struct {
+		configPath  string
+		engineState string
+	}{
+		{
+			configPath:  filepath.Join(root, defaultConfigFilename),
+			engineState: filepath.Join(root, "state.json"),
+		},
+		{
+			configPath:  filepath.Join(root, profilesSubdir, "abc123.json"),
+			engineState: filepath.Join(root, profilesSubdir, "abc123.state.json"),
+		},
+	}
+
+	for _, c := range cases {
+		account, err := profileAccountPathFor(c.configPath)
+		if err != nil {
+			t.Fatalf("%s: %v", c.configPath, err)
+		}
+		if account == c.engineState {
+			t.Errorf("account file collides with the engine state file: %q", account)
+		}
+	}
+}
+
+func TestWriteThenReadProfileEmail(t *testing.T) {
+	configPath := filepath.Join(t.TempDir(), "profiles", "abc123.json")
+	if err := ensureDirFor(t, configPath); err != nil {
+		t.Fatalf("prepare dir: %v", err)
+	}
+
+	if got := readProfileEmail(configPath); got != "" {
+		t.Errorf("expected no email before a login, got %q", got)
+	}
+
+	const email = "user@example.com"
+	if err := writeProfileEmail(configPath, email); err != nil {
+		t.Fatalf("write: %v", err)
+	}
+
+	if got := readProfileEmail(configPath); got != email {
+		t.Errorf("got %q, want %q", got, email)
+	}
+
+	if err := removeProfileEmail(configPath); err != nil {
+		t.Fatalf("remove: %v", err)
+	}
+	if got := readProfileEmail(configPath); got != "" {
+		t.Errorf("expected no email after logout, got %q", got)
+	}
+
+	// Logout may run on a never-logged-in profile, so a second remove must pass.
+	if err := removeProfileEmail(configPath); err != nil {
+		t.Fatalf("second remove should be a no-op: %v", err)
+	}
+}
+
+func TestWriteProfileEmailIgnoresEmpty(t *testing.T) {
+	configPath := filepath.Join(t.TempDir(), "profiles", "abc123.json")
+	if err := ensureDirFor(t, configPath); err != nil {
+		t.Fatalf("prepare dir: %v", err)
+	}
+
+	const email = "user@example.com"
+	if err := writeProfileEmail(configPath, email); err != nil {
+		t.Fatalf("write: %v", err)
+	}
+	if err := writeProfileEmail(configPath, ""); err != nil {
+		t.Fatalf("write empty: %v", err)
+	}
+
+	if got := readProfileEmail(configPath); got != email {
+		t.Errorf("empty write clobbered the stored email: got %q, want %q", got, email)
+	}
+}
+
+func ensureDirFor(t *testing.T, path string) error {
+	t.Helper()
+	return os.MkdirAll(filepath.Dir(path), 0o700)
+}

client/android/session.go

@@ -278,7 +278,7 @@ func (c *Client) endExtend() {
 }
 
 func (c *Client) extendAuthSession(ctx context.Context, urlOpener URLOpener, isAndroidTV bool) error {
-	cfg, _, cc := c.stateSnapshot()
+	cfg, cfgPath, cc := c.authSnapshot()
 	if cfg == nil || cc == nil {
 		return fmt.Errorf("engine is not running")
 	}
@@ -293,7 +293,10 @@ func (c *Client) extendAuthSession(ctx context.Context, urlOpener URLOpener, isA
 	}
 	defer authClient.Close()
 
-	a := &Auth{ctx: ctx, config: cfg}
+	// Passing the config path makes the flow pick up the login_hint: an extend
+	// renews the session of the account already signed in, so it must not stop to
+	// offer a choice.
+	a := NewAuthWithConfig(ctx, cfg, cfgPath)
 	tokenInfo, err := a.foregroundGetTokenInfo(authClient, urlOpener, isAndroidTV)
 	if err != nil {
 		return fmt.Errorf("interactive sso login failed: %v", err)

client/embed/embed.go

@@ -85,11 +85,6 @@ type Options struct {
 	DisableIPv6 bool
 	// BlockInbound blocks all inbound connections from peers
 	BlockInbound bool
-	// EnableRosenpass enables the Rosenpass post-quantum key exchange.
-	EnableRosenpass bool
-	// RosenpassPermissive lets a Rosenpass-enabled peer still connect to peers
-	// that do not run Rosenpass (falling back to the plain WireGuard PSK).
-	RosenpassPermissive bool
 	// BlockLANAccess blocks the embedded peer from reaching the host's
 	// LAN (RFC 1918, link-local, loopback) when it's used as a routing
 	// peer. Mirrors profilemanager.ConfigInput.BlockLANAccess. Useful
@@ -208,8 +203,6 @@ func New(opts Options) (*Client, error) {
 		DisableIPv6:         &opts.DisableIPv6,
 		BlockInbound:        &opts.BlockInbound,
 		BlockLANAccess:      &opts.BlockLANAccess,
-		RosenpassEnabled:    &opts.EnableRosenpass,
-		RosenpassPermissive: &opts.RosenpassPermissive,
 		WireguardPort:       opts.WireguardPort,
 		MTU:                 opts.MTU,
 		DNSLabels:           parsedLabels,

client/internal/conn_mgr.go

@@ -2,7 +2,6 @@ package internal
 
 import (
 	"context"
-	"maps"
 	"os"
 	"strconv"
 	"sync"
@@ -15,7 +14,6 @@ import (
 	"github.com/netbirdio/netbird/client/internal/peer"
 	"github.com/netbirdio/netbird/client/internal/peerstore"
 	"github.com/netbirdio/netbird/route"
-	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
 )
 
 // lazyForce is the resolved local decision for lazy connections, layered above the
@@ -39,13 +37,11 @@ const (
 // The only exception is ActivatePeer, which is safe for concurrent use so the
 // DNS warm-up path can call it without contending on the engine mutex.
 type ConnMgr struct {
-	peerStore      *peerstore.Store
-	statusRecorder *peer.Status
-	iface          lazyconn.WGIface
-	force          lazyForce
-	// remoteLazyEnabled caches the account-wide lazy feature flag from management.
-	// It is the default for peers that do not carry a per-peer lazy hint.
-	remoteLazyEnabled bool
+	peerStore        *peerstore.Store
+	statusRecorder   *peer.Status
+	iface            lazyconn.WGIface
+	force            lazyForce
+	rosenpassEnabled bool
 
 	lazyConnMgr *manager.Manager
 	// lazyConnMgrMu guards the lazyConnMgr pointer for readers outside the
@@ -57,10 +53,6 @@ type ConnMgr struct {
 	// (re)armed (Mode A at arm time). Injected by the engine; nil disables the reconcile.
 	reconcileRoutedIPs func(peerKey string) error
 
-	// appliedExcludeList is the exclude set last handed to the lazy manager, kept so an
-	// unchanged set on the next sync skips the O(n) reconciliation.
-	appliedExcludeList map[string]bool
-
 	wg            sync.WaitGroup
 	lazyCtx       context.Context
 	lazyCtxCancel context.CancelFunc
@@ -74,59 +66,78 @@ func (e *ConnMgr) SetRoutedIPsReconciler(fn func(peerKey string) error) {
 
 func NewConnMgr(engineConfig *EngineConfig, statusRecorder *peer.Status, peerStore *peerstore.Store, iface lazyconn.WGIface) *ConnMgr {
 	e := &ConnMgr{
-		peerStore:      peerStore,
-		statusRecorder: statusRecorder,
-		iface:          iface,
-		force:          resolveLazyForce(engineConfig.LazyConnection),
+		peerStore:        peerStore,
+		statusRecorder:   statusRecorder,
+		iface:            iface,
+		force:            resolveLazyForce(engineConfig.LazyConnection),
+		rosenpassEnabled: engineConfig.RosenpassEnabled,
 	}
 	return e
 }
 
-// Start initializes the connection manager. The lazy connection manager always runs so that
-// per-peer lazy defaults (e.g. proxy peers) work even when the account flag is off; the
-// account flag and the local override decide the default lazy state per peer (see
-// PeerLazyDefault). Rosenpass peers stay lazy-capable too: their connections just never idle
-// on their own, since rosenpass rekey traffic keeps them active.
+// Start initializes the connection manager. It starts the lazy connection manager when a
+// local override forces it on; with no local override it waits for the management feature flag.
 func (e *ConnMgr) Start(ctx context.Context) {
 	if e.lazyConnMgr != nil {
 		log.Errorf("lazy connection manager is already started")
 		return
 	}
 
-	e.initLazyManager(ctx)
-	e.statusRecorder.UpdateLazyConnection(e.PeerLazyDefault(mgmProto.LazyState_LazyStateDefault))
-}
-
-// UpdatedRemoteFeatureFlag caches the account-wide lazy feature flag. The manager itself is
-// not started or stopped here; the per-sync exclude-list reconciliation moves normal peers
-// between the lazy and always-active sets when the flag flips.
-func (e *ConnMgr) UpdatedRemoteFeatureFlag(_ context.Context, enabled bool) error {
-	e.remoteLazyEnabled = enabled
-	if e.isStartedWithLazyMgr() {
-		e.statusRecorder.UpdateLazyConnection(e.PeerLazyDefault(mgmProto.LazyState_LazyStateDefault))
-	}
-	return nil
-}
-
-// PeerLazyDefault reports whether a peer should be lazy. The local override
-// (NB_LAZY_CONN/MDM) wins over everything; without a local override the
-// management per-peer state applies (LazyStateLazy/Eager force the decision),
-// and LazyStateDefault follows the account-wide flag.
-func (e *ConnMgr) PeerLazyDefault(state mgmProto.LazyState) bool {
 	switch e.force {
-	case lazyForceOn:
-		return true
 	case lazyForceOff:
-		return false
+		log.Infof("lazy connection manager is disabled by local override (%s or MDM policy)", lazyconn.EnvLazyConn)
+		e.statusRecorder.UpdateLazyConnection(false)
+		return
+	case lazyForceNone:
+		log.Infof("lazy connection manager is managed by the management feature flag")
+		e.statusRecorder.UpdateLazyConnection(false)
+		return
 	}
 
-	switch state {
-	case mgmProto.LazyState_LazyStateLazy:
-		return true
-	case mgmProto.LazyState_LazyStateEager:
-		return false
-	default:
-		return e.remoteLazyEnabled
+	if e.rosenpassEnabled {
+		log.Warnf("rosenpass connection manager is enabled, lazy connection manager will not be started")
+		e.statusRecorder.UpdateLazyConnection(false)
+		return
+	}
+
+	e.initLazyManager(ctx)
+	e.statusRecorder.UpdateLazyConnection(true)
+}
+
+// UpdatedRemoteFeatureFlag is called when the remote feature flag is updated.
+// If enabled, it initializes the lazy connection manager and start it. Do not need to call Start() again.
+// If disabled, then it closes the lazy connection manager and open the connections to all peers.
+func (e *ConnMgr) UpdatedRemoteFeatureFlag(ctx context.Context, enabled bool) error {
+	// a local override (NB_LAZY_CONN or local config) takes precedence over management
+	if e.force != lazyForceNone {
+		return nil
+	}
+
+	if enabled {
+		// if the lazy connection manager is already started, do not start it again
+		if e.lazyConnMgr != nil {
+			return nil
+		}
+
+		if e.rosenpassEnabled {
+			log.Infof("rosenpass connection manager is enabled, lazy connection manager will not be started")
+			e.statusRecorder.UpdateLazyConnection(false)
+			return nil
+		}
+
+		log.Infof("lazy connection manager is enabled by the management feature flag")
+		e.initLazyManager(ctx)
+		e.statusRecorder.UpdateLazyConnection(true)
+		return e.addPeersToLazyConnManager()
+	} else {
+		if e.lazyConnMgr == nil {
+			e.statusRecorder.UpdateLazyConnection(false)
+			return nil
+		}
+		log.Infof("lazy connection manager is disabled by management feature flag")
+		e.closeManager(ctx)
+		e.statusRecorder.UpdateLazyConnection(false)
+		return nil
 	}
 }
 
@@ -146,13 +157,6 @@ func (e *ConnMgr) SetExcludeList(ctx context.Context, peerIDs map[string]bool) {
 		return
 	}
 
-	// The exclude set is recomputed every sync but rarely changes; skip the O(n)
-	// store lookups and reconciliation when it matches what was already applied.
-	if maps.Equal(peerIDs, e.appliedExcludeList) {
-		return
-	}
-	e.appliedExcludeList = maps.Clone(peerIDs)
-
 	excludedPeers := make([]lazyconn.PeerConfig, 0, len(peerIDs))
 
 	for peerID := range peerIDs {
@@ -188,16 +192,12 @@ func (e *ConnMgr) SetExcludeList(ctx context.Context, peerIDs map[string]bool) {
 	}
 }
 
-// AddPeerConn registers a peer connection. permanent requests an always-active connection
-// (the peer belongs to the exclude set: a forwarder, or a peer that is not lazy by policy).
-// Non-permanent peers are handed to the lazy manager. The subsequent SetExcludeList call
-// reconciles membership for existing peers across flag flips.
-func (e *ConnMgr) AddPeerConn(ctx context.Context, peerKey string, conn *peer.Conn, permanent bool) (exists bool) {
+func (e *ConnMgr) AddPeerConn(ctx context.Context, peerKey string, conn *peer.Conn) (exists bool) {
 	if success := e.peerStore.AddPeerConn(peerKey, conn); !success {
 		return true
 	}
 
-	if !e.isStartedWithLazyMgr() || permanent {
+	if !e.isStartedWithLazyMgr() {
 		if err := conn.Open(ctx); err != nil {
 			conn.Log.Errorf("failed to open connection: %v", err)
 		}
@@ -296,8 +296,6 @@ func (e *ConnMgr) Close() {
 	e.lazyConnMgrMu.Lock()
 	e.lazyConnMgr = nil
 	e.lazyConnMgrMu.Unlock()
-
-	e.appliedExcludeList = nil
 }
 
 func (e *ConnMgr) initLazyManager(engineCtx context.Context) {
@@ -311,8 +309,6 @@ func (e *ConnMgr) initLazyManager(engineCtx context.Context) {
 	e.lazyCtx, e.lazyCtxCancel = context.WithCancel(engineCtx)
 	e.lazyConnMgrMu.Unlock()
 
-	e.appliedExcludeList = nil
-
 	e.wg.Add(1)
 	go func() {
 		defer e.wg.Done()
@@ -320,6 +316,46 @@ func (e *ConnMgr) initLazyManager(engineCtx context.Context) {
 	}()
 }
 
+func (e *ConnMgr) addPeersToLazyConnManager() error {
+	peers := e.peerStore.PeersPubKey()
+	lazyPeerCfgs := make([]lazyconn.PeerConfig, 0, len(peers))
+	for _, peerID := range peers {
+		var peerConn *peer.Conn
+		var exists bool
+		if peerConn, exists = e.peerStore.PeerConn(peerID); !exists {
+			log.Warnf("failed to find peer conn for peerID: %s", peerID)
+			continue
+		}
+
+		lazyPeerCfg := lazyconn.PeerConfig{
+			PublicKey:  peerID,
+			AllowedIPs: peerConn.WgConfig().AllowedIps,
+			PeerConnID: peerConn.ConnID(),
+			Log:        peerConn.Log,
+		}
+		lazyPeerCfgs = append(lazyPeerCfgs, lazyPeerCfg)
+	}
+
+	return e.lazyConnMgr.AddActivePeers(lazyPeerCfgs)
+}
+
+func (e *ConnMgr) closeManager(ctx context.Context) {
+	if e.lazyConnMgr == nil {
+		return
+	}
+
+	e.lazyCtxCancel()
+	e.wg.Wait()
+
+	e.lazyConnMgrMu.Lock()
+	e.lazyConnMgr = nil
+	e.lazyConnMgrMu.Unlock()
+
+	for _, peerID := range e.peerStore.PeersPubKey() {
+		e.peerStore.PeerConnOpen(ctx, peerID)
+	}
+}
+
 func (e *ConnMgr) isStartedWithLazyMgr() bool {
 	return e.lazyConnMgr != nil && e.lazyCtxCancel != nil
 }

client/internal/conn_mgr_test.go

@@ -16,7 +16,6 @@ import (
 	"github.com/netbirdio/netbird/client/internal/peer"
 	"github.com/netbirdio/netbird/client/internal/peerstore"
 	"github.com/netbirdio/netbird/monotime"
-	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
 )
 
 func TestResolveLazyForce(t *testing.T) {
@@ -139,91 +138,4 @@ func TestInactivityThresholdEnv(t *testing.T) {
 	}
 }
 
-func TestPeerLazyDefault(t *testing.T) {
-	tests := []struct {
-		name          string
-		force         lazyForce
-		remoteEnabled bool
-		state         mgmProto.LazyState
-		want          bool
-	}{
-		{name: "force on wins over eager state", force: lazyForceOn, state: mgmProto.LazyState_LazyStateEager, want: true},
-		{name: "force off wins over lazy state", force: lazyForceOff, remoteEnabled: true, state: mgmProto.LazyState_LazyStateLazy, want: false},
-		{name: "none, default, account off -> active", force: lazyForceNone, state: mgmProto.LazyState_LazyStateDefault, want: false},
-		{name: "none, default, account on -> lazy", force: lazyForceNone, remoteEnabled: true, state: mgmProto.LazyState_LazyStateDefault, want: true},
-		{name: "none, lazy state, account off -> lazy", force: lazyForceNone, state: mgmProto.LazyState_LazyStateLazy, want: true},
-		{name: "none, eager state, account on -> active", force: lazyForceNone, remoteEnabled: true, state: mgmProto.LazyState_LazyStateEager, want: false},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			e := &ConnMgr{force: tt.force, remoteLazyEnabled: tt.remoteEnabled}
-			if got := e.PeerLazyDefault(tt.state); got != tt.want {
-				t.Fatalf("PeerLazyDefault(%v) = %v, want %v", tt.state, got, tt.want)
-			}
-		})
-	}
-}
-
 func durPtr(d time.Duration) *time.Duration { return &d }
-
-// TestToExcludedLazyPeers covers the per-peer lazy classification (proxy vs
-// normal, across the force/account-flag matrix). Forwarder-target exclusion is
-// covered by TestToExcludedLazyPeers_ForwardTarget.
-func TestToExcludedLazyPeers(t *testing.T) {
-	const (
-		normalKey = "normal"
-		lazyKey   = "lazy-state"
-		eagerKey  = "eager-state"
-	)
-
-	peers := []*mgmProto.RemotePeerConfig{
-		{WgPubKey: normalKey, AllowedIps: []string{"100.64.0.1/32"}},
-		{WgPubKey: lazyKey, AllowedIps: []string{"100.64.0.2/32"}, LazyState: mgmProto.LazyState_LazyStateLazy},
-		{WgPubKey: eagerKey, AllowedIps: []string{"100.64.0.3/32"}, LazyState: mgmProto.LazyState_LazyStateEager},
-	}
-
-	tests := []struct {
-		name          string
-		force         lazyForce
-		remoteEnabled bool
-		want          map[string]bool
-	}{
-		{
-			name:  "account off: lazy-state peer lazy, normal + eager active",
-			force: lazyForceNone, remoteEnabled: false,
-			want: map[string]bool{normalKey: true, eagerKey: true},
-		},
-		{
-			name:  "account on: only eager-state peer active",
-			force: lazyForceNone, remoteEnabled: true,
-			want: map[string]bool{eagerKey: true},
-		},
-		{
-			name:  "force off: everything active",
-			force: lazyForceOff, remoteEnabled: true,
-			want: map[string]bool{normalKey: true, lazyKey: true, eagerKey: true},
-		},
-		{
-			name:  "force on: nothing active",
-			force: lazyForceOn, remoteEnabled: false,
-			want: map[string]bool{},
-		},
-	}
-
-	for _, tt := range tests {
-		t.Run(tt.name, func(t *testing.T) {
-			e := &Engine{connMgr: &ConnMgr{force: tt.force, remoteLazyEnabled: tt.remoteEnabled}}
-			got := e.toExcludedLazyPeers(peers)
-
-			if len(got) != len(tt.want) {
-				t.Fatalf("toExcludedLazyPeers() = %v, want %v", got, tt.want)
-			}
-			for k := range tt.want {
-				if !got[k] {
-					t.Fatalf("expected peer %s excluded, got %v", k, got)
-				}
-			}
-		})
-	}
-}

client/internal/connect.go

@@ -113,11 +113,14 @@ func (c *ConnectClient) RunOnAndroid(
 	stateFilePath string,
 	cacheDir string,
 ) error {
+	notifier := tunnelnotifier.New(networkChangeListener, nil)
+	defer notifier.Close()
+
 	// in case of non Android os these variables will be nil
 	mobileDependency := MobileDependency{
 		TunAdapter:            tunAdapter,
 		IFaceDiscover:         iFaceDiscover,
-		NetworkChangeListener: networkChangeListener,
+		NetworkChangeListener: notifier,
 		HostDNSAddresses:      dnsAddresses,
 		DnsReadyListener:      dnsReadyListener,
 		StateFilePath:         stateFilePath,

client/internal/daemonaddr/identity.go

@@ -0,0 +1,14 @@
+package daemonaddr
+
+import "strings"
+
+// CarriesIdentity reports whether the control channel at addr conveys the
+// connecting process's identity to the daemon. A Unix socket carries peer
+// credentials and a named pipe carries the client's token; loopback TCP carries
+// neither, so on such an address the daemon cannot authorize a privileged
+// operation for anybody. A client uses this to tell whether becoming privileged
+// would get it anywhere: on an identity-less address it would not, and the only
+// way forward is to move the daemon onto one that carries identity.
+func CarriesIdentity(addr string) bool {
+	return strings.HasPrefix(addr, "unix://") || strings.HasPrefix(addr, pipeScheme)
+}

client/internal/daemonaddr/identity_test.go

@@ -0,0 +1,29 @@
+package daemonaddr
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestCarriesIdentity(t *testing.T) {
+	tests := []struct {
+		addr string
+		want bool
+	}{
+		{"unix:///var/run/netbird.sock", true},
+		{"unix:///var/run/netbird/default.sock", true},
+		{"npipe://netbird", true},
+		{`npipe://\\.\pipe\ProtectedPrefix\Administrators\netbird`, true},
+		{"tcp://127.0.0.1:41731", false},
+		{"tcp://localhost:41731", false},
+		{"", false},
+		{"/var/run/netbird.sock", false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.addr, func(t *testing.T) {
+			assert.Equal(t, tt.want, CarriesIdentity(tt.addr), "address %q", tt.addr)
+		})
+	}
+}

client/internal/dns/interface_index.go

@@ -0,0 +1,15 @@
+package dns
+
+import (
+	"fmt"
+	"net"
+)
+
+func getInterfaceIndex(interfaceName string) (int, error) {
+	iface, err := net.InterfaceByName(interfaceName)
+	if err != nil {
+		return 0, fmt.Errorf("lookup interface %q: %w", interfaceName, err)
+	}
+
+	return iface.Index, nil
+}

client/internal/dns/interface_index_test.go

@@ -0,0 +1,35 @@
+package dns
+
+import (
+	"net"
+	"testing"
+)
+
+func TestGetInterfaceIndexExisting(t *testing.T) {
+	interfaces, err := net.Interfaces()
+	if err != nil {
+		t.Fatalf("list network interfaces: %v", err)
+	}
+	if len(interfaces) == 0 {
+		t.Fatal("expected at least one network interface")
+	}
+
+	iface := interfaces[0]
+	index, err := getInterfaceIndex(iface.Name)
+	if err != nil {
+		t.Fatalf("look up existing interface %q: %v", iface.Name, err)
+	}
+	if index != iface.Index {
+		t.Fatalf("expected interface index %d, got %d", iface.Index, index)
+	}
+}
+
+func TestGetInterfaceIndexMissing(t *testing.T) {
+	index, err := getInterfaceIndex("netbird-interface-that-does-not-exist")
+	if index != 0 {
+		t.Fatalf("expected missing interface index to be 0, got %d", index)
+	}
+	if err == nil {
+		t.Fatal("expected missing interface lookup to return an error")
+	}
+}

client/internal/dns/notifier.go

@@ -51,7 +51,5 @@ func (n *notifier) notify() {
 		return
 	}
 
-	go func(l listener.NetworkChangeListener) {
-		l.OnNetworkChanged("")
-	}(n.listener)
+	n.listener.OnNetworkChanged("")
 }

client/internal/dns/server.go

@@ -252,7 +252,7 @@ func NewDefaultServerPermanentUpstream(
 	ds.hostsDNSHolder.set(hostsDnsList)
 	ds.permanent = true
 	ds.currentConfig = dnsConfigToHostDNSConfig(config, ds.service.RuntimeIP(), ds.service.RuntimePort())
-	ds.searchDomainNotifier = newNotifier(ds.SearchDomains())
+	ds.searchDomainNotifier = newNotifier(ds.searchDomains())
 	ds.searchDomainNotifier.setListener(listener)
 	setServerDns(ds)
 	return ds
@@ -602,6 +602,12 @@ func (s *DefaultServer) UpdateDNSServer(serial uint64, update nbdns.Config) erro
 }
 
 func (s *DefaultServer) SearchDomains() []string {
+	s.mux.Lock()
+	defer s.mux.Unlock()
+	return s.searchDomains()
+}
+
+func (s *DefaultServer) searchDomains() []string {
 	var searchDomains []string
 
 	for _, dConf := range s.currentConfig.Domains {
@@ -686,7 +692,7 @@ func (s *DefaultServer) applyConfiguration(update nbdns.Config) error {
 	}()
 
 	if s.searchDomainNotifier != nil {
-		s.searchDomainNotifier.onNewSearchDomains(s.SearchDomains())
+		s.searchDomainNotifier.onNewSearchDomains(s.searchDomains())
 	}
 
 	s.updateNSGroupStates(update.NameServerGroups)

client/internal/dns/upstream_ios.go

@@ -130,8 +130,3 @@ func GetClientPrivate(iface privateClientIface, upstreamIP netip.Addr, dialTimeo
 	}
 	return client, nil
 }
-
-func getInterfaceIndex(interfaceName string) (int, error) {
-	iface, err := net.InterfaceByName(interfaceName)
-	return iface.Index, err
-}

client/internal/elevate/elevate.go

@@ -0,0 +1,74 @@
+// Package elevate re-runs this very executable under the operating system's own
+// privilege-elevation mechanism and waits for it to finish.
+//
+// It exists so that a change the daemon restricts to root/administrator can be
+// authorized from the GUI, by the user, at the moment they ask for it: Windows
+// shows the UAC consent dialog, macOS the system authentication dialog, and
+// Linux/FreeBSD the session's polkit agent. The credentials, where any are
+// asked for, are collected by the operating system and never pass through
+// NetBird.
+//
+// What the elevated process then does is the caller's business: it is the same
+// binary, in a one-shot mode, and it is authorized by the daemon exactly like
+// any other privileged caller, from the identity the kernel reports on the
+// control channel. Nothing here grants privilege, and the daemon gains no new
+// way to be talked into something: elevation only changes who is calling it.
+package elevate
+
+import (
+	"context"
+	"errors"
+
+	log "github.com/sirupsen/logrus"
+)
+
+// AppliedMarker is what the elevated process prints on standard output once it has
+// done what it was run for.
+//
+// macOS's AuthorizationExecuteWithPrivileges reports no exit status and does not
+// say which process it started, so there this line is the only evidence that the
+// change was applied. The other platforms have an exit code and ignore it.
+const AppliedMarker = "netbird-elevated: applied"
+
+var (
+	// ErrDeclined reports that the user dismissed the prompt or did not
+	// authenticate. Nothing happened and nothing is wrong: a caller undoes its
+	// optimistic update and stays quiet.
+	ErrDeclined = errors.New("authorization declined")
+
+	// ErrUnavailable reports that this host has no elevation mechanism we can
+	// drive: no polkit on a Unix desktop, or an executable we decline to run as
+	// root. A caller falls back to telling the user which command to run.
+	ErrUnavailable = errors.New("no privilege elevation mechanism available")
+)
+
+// Run runs this executable with args under the platform's elevation mechanism
+// and waits for it to exit. A non-zero exit is returned as an error, so the
+// caller can treat a completed Run as the operation having succeeded.
+//
+// The args are the caller's own command line, so they cross no privilege
+// boundary: only a user who has just authenticated as an administrator can get
+// them run at all.
+func Run(ctx context.Context, args ...string) error {
+	self, err := trustedSelf()
+	if err != nil {
+		return err
+	}
+	return run(ctx, self, args)
+}
+
+// Available reports whether Run has a mechanism to use on this host, so a caller
+// can offer the prompt only when there is one and otherwise fall back to
+// guidance the user can act on. It answers from what is installed, not from what
+// the user is allowed to do: an administrator's password may still be required
+// and may still not be given, which is ErrDeclined from Run.
+func Available() bool {
+	if _, err := trustedSelf(); err != nil {
+		// Worth a line: this is also what a build run from a group-writable
+		// directory hits, and there is nothing in the UI to say why the offer is
+		// missing.
+		log.Debugf("not offering privilege elevation: %v", err)
+		return false
+	}
+	return mechanismAvailable()
+}

client/internal/elevate/output.go

@@ -0,0 +1,19 @@
+package elevate
+
+import "strings"
+
+// noOutput stands in for a process that said nothing, so that a report of what it
+// said still reads as a sentence.
+const noOutput = "no output"
+
+// firstLine trims captured output to something that reads in one line.
+func firstLine(s string) string {
+	s = strings.TrimSpace(s)
+	if s == "" {
+		return noOutput
+	}
+	if i := strings.IndexByte(s, '\n'); i >= 0 {
+		return s[:i]
+	}
+	return s
+}

client/internal/elevate/output_test.go

@@ -0,0 +1,21 @@
+package elevate
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestFirstLine(t *testing.T) {
+	tests := []struct{ in, want string }{
+		{in: "", want: noOutput},
+		{in: "  \n ", want: noOutput},
+		{in: "one line", want: "one line"},
+		{in: "first\nsecond", want: "first"},
+		{in: "\nsecond\n", want: "second"},
+	}
+
+	for _, tt := range tests {
+		assert.Equal(t, tt.want, firstLine(tt.in), "input %q", tt.in)
+	}
+}

client/internal/elevate/run_darwin.go

@@ -0,0 +1,359 @@
+package elevate
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"os"
+	"runtime"
+	"strings"
+	"sync"
+	"syscall"
+	"unsafe"
+
+	"github.com/ebitengine/purego"
+	log "github.com/sirupsen/logrus"
+)
+
+// Authorization Services, reached through purego rather than cgo so the released
+// binaries keep building with CGO_ENABLED=0.
+//
+// The prompt belongs to this process, which is what makes it carry the
+// application's name and our own explanation. Going through osascript instead puts
+// the very same trampoline behind a dialog attributed to osascript, and means
+// handing a shell a command line to re-parse.
+//
+// # On AuthorizationExecuteWithPrivileges
+//
+// It is deprecated, and Apple's guidance (Quinn, "BSD Privilege Escalation on
+// macOS", developer.apple.com/forums/thread/708765) is "while it still works, it's
+// been deprecated for many years. Do not use it in a widely distributed product."
+// It is used here anyway, knowingly, because the alternatives Apple offers are for
+// *obtaining* ongoing privileges — an installer package, SMAppService, SMJobBless —
+// and NetBird already has what they would install: a launchd daemon running as
+// root. What is missing is only a way for an unprivileged client to ask it to act.
+//
+// The way to that without a deprecated call is to authorize the client instead of
+// elevating one: the app takes the right with AuthorizationCreate, passes the
+// AuthorizationExternalForm to the daemon, and the daemon checks it with
+// AuthorizationCopyRights before acting — none of which is deprecated. It is the
+// better design and it is where this should end up. It also means the daemon
+// accepting an authorization over its control socket, which is a new way to be
+// asked for privileged work and wants reviewing as such, so it is deliberately not
+// bundled in with the rest of this.
+//
+// Until then, three things keep the deprecation from being a trap. Every symbol is
+// resolved with an error rather than a panic, so a macOS that has dropped this
+// function leaves the app offering the user a command instead of crashing on the
+// way to a prompt. A failure to run the tool is reported as ErrUnavailable, so the
+// fallback is the same one an agent-less Linux session gets. And the whole path
+// runs under guard, which turns a panic out of the FFI layer into that same
+// fallback.
+//
+// The trampoline passes on the environment it was given, so what it starts as root
+// must be an executable this user's peers cannot influence: that is what
+// trustedSelf refuses, and what signing the binary settles for the loader.
+
+const (
+	securityFramework = "/System/Library/Frameworks/Security.framework/Security"
+	libSystem         = "/usr/lib/libSystem.B.dylib"
+
+	// trampoline is what the framework hands the tool to. Present on every macOS,
+	// and worth confirming before offering a prompt rather than mid-prompt.
+	trampoline = "/usr/libexec/security_authtrampoline"
+)
+
+// rightExecute is the right an administrator holds, and what
+// AuthorizationExecuteWithPrivileges requires of us.
+const rightExecute = "system.privilege.admin"
+
+// promptKey is kAuthorizationEnvironmentPrompt, which puts a sentence of ours above
+// the system's in the dialog. It is about the change rather than the mechanism.
+const (
+	promptKey  = "prompt"
+	promptText = "NetBird needs to change a setting that grants SSH access to this computer."
+)
+
+// OSStatus values from SecBase.h that mean something to us; anything else is
+// reported as it comes.
+const (
+	errAuthorizationSuccess               = 0
+	errAuthorizationDenied                = -60005
+	errAuthorizationCanceled              = -60006
+	errAuthorizationInteractionNotAllowed = -60007
+	errAuthorizationToolExecuteFailure    = -60031
+	errAuthorizationToolEnvironmentError  = -60032
+)
+
+// AuthorizationFlags from Authorization.h.
+const (
+	flagDefaults           = 0
+	flagInteractionAllowed = 1 << 0
+	flagExtendRights       = 1 << 1
+	flagDestroyRights      = 1 << 3
+	flagPreAuthorize       = 1 << 4
+)
+
+// authorizationItem mirrors AuthorizationItem: a name, and a value the name gives
+// meaning to. 32 bytes on both amd64 and arm64.
+type authorizationItem struct {
+	name        *byte
+	valueLength uintptr
+	value       unsafe.Pointer
+	// flags is reserved by the API and always zero. Declared because the layout
+	// is the contract: without it the struct is 24 bytes where C reads 32.
+	flags uint32 //nolint:unused // part of the C layout
+}
+
+// authorizationItemSet mirrors AuthorizationItemSet, which serves as both an
+// AuthorizationRights and an AuthorizationEnvironment.
+type authorizationItemSet struct {
+	count uint32
+	items *authorizationItem
+}
+
+var (
+	authorizationCreate                func(rights, environment *authorizationItemSet, flags uint32, authorization *uintptr) int32
+	authorizationExecuteWithPrivileges func(authorization uintptr, pathToTool string, options uint32, arguments *uintptr, communicationsPipe *uintptr) int32
+	authorizationFree                  func(authorization uintptr, flags uint32) int32
+	fileno                             func(stream uintptr) int32
+	fclose                             func(stream uintptr) int32
+
+	loadOnce sync.Once
+	loadErr  error
+)
+
+// load resolves the functions once. A framework that cannot be opened, or a symbol
+// that is no longer there, leaves the host without a mechanism rather than taking
+// the process down with it: see the note on deprecation above.
+func load() error {
+	loadOnce.Do(func() { loadErr = guard("loading Security.framework", resolve) })
+	return loadErr
+}
+
+// guard turns a panic out of the FFI layer into an error, so an API that has
+// changed under us costs the user a prompt rather than the window they were
+// clicking in. purego panics on a signature it cannot map, and this is the one
+// place in the client that calls a deprecated system function.
+//
+// It catches Go panics, which is what purego raises. A fault inside the framework
+// itself is not a panic and not recoverable; the layout the tests pin down is what
+// stands between us and that.
+func guard(what string, fn func() error) (err error) {
+	defer func() {
+		r := recover()
+		if r == nil {
+			return
+		}
+		log.Errorf("%s panicked: %v", what, r)
+		err = fmt.Errorf("%w: %s: %v", ErrUnavailable, what, r)
+	}()
+	return fn()
+}
+
+func resolve() error {
+	security, err := purego.Dlopen(securityFramework, purego.RTLD_LAZY|purego.RTLD_GLOBAL)
+	if err != nil {
+		return fmt.Errorf("open %s: %w", securityFramework, err)
+	}
+	system, err := purego.Dlopen(libSystem, purego.RTLD_LAZY|purego.RTLD_GLOBAL)
+	if err != nil {
+		return fmt.Errorf("open %s: %w", libSystem, err)
+	}
+
+	// purego.RegisterLibFunc panics on a symbol it cannot find, which is not how a
+	// deprecated function's disappearance should reach the user.
+	for _, fn := range []struct {
+		ptr    any
+		handle uintptr
+		name   string
+	}{
+		{&authorizationCreate, security, "AuthorizationCreate"},
+		{&authorizationExecuteWithPrivileges, security, "AuthorizationExecuteWithPrivileges"},
+		{&authorizationFree, security, "AuthorizationFree"},
+		{&fileno, system, "fileno"},
+		{&fclose, system, "fclose"},
+	} {
+		symbol, err := purego.Dlsym(fn.handle, fn.name)
+		if err != nil {
+			return fmt.Errorf("resolve %s: %w", fn.name, err)
+		}
+		if symbol == 0 {
+			return fmt.Errorf("resolve %s: not present on this system", fn.name)
+		}
+		purego.RegisterFunc(fn.ptr, symbol)
+	}
+	return nil
+}
+
+// run asks the system to run self as root: first for the right, which is what puts
+// up the authentication dialog and collects the password or takes the Touch ID,
+// then for the tool. The credentials go to the system's authorization trampoline
+// and never to us.
+//
+// The context bounds only our own waiting; the dialog belongs to the system and
+// closes when the user answers it.
+func run(ctx context.Context, self string, args []string) error {
+	if err := load(); err != nil {
+		return fmt.Errorf("%w: %v", ErrUnavailable, err)
+	}
+
+	return guard("asking for privileges", func() error {
+		authorization, err := authorize()
+		if err != nil {
+			return err
+		}
+		defer authorizationFree(authorization, flagDestroyRights)
+
+		return execute(ctx, authorization, self, args)
+	})
+}
+
+func mechanismAvailable() bool {
+	if err := load(); err != nil {
+		return false
+	}
+	info, err := os.Stat(trampoline)
+	return err == nil && !info.IsDir()
+}
+
+// authorize obtains the right, prompting for it. A dismissed dialog comes back as
+// errAuthorizationCanceled and a password given up on as errAuthorizationDenied;
+// both are the user's answer rather than a failure.
+func authorize() (uintptr, error) {
+	var pinner runtime.Pinner
+	defer pinner.Unpin()
+
+	rights := itemSet(&pinner, authorizationItem{name: cString(&pinner, rightExecute)})
+	environment := itemSet(&pinner, promptItem(&pinner))
+
+	var authorization uintptr
+	status := authorizationCreate(rights, environment,
+		flagDefaults|flagInteractionAllowed|flagPreAuthorize|flagExtendRights, &authorization)
+
+	switch status {
+	case errAuthorizationSuccess:
+		return authorization, nil
+	case errAuthorizationCanceled, errAuthorizationDenied:
+		return 0, ErrDeclined
+	case errAuthorizationInteractionNotAllowed:
+		// Nowhere to put a dialog, so there is nobody to ask: a launch daemon, or
+		// a session with no window server.
+		return 0, fmt.Errorf("%w: this session cannot show an authorization prompt", ErrUnavailable)
+	default:
+		return 0, fmt.Errorf("request %s: OSStatus %d", rightExecute, status)
+	}
+}
+
+// execute runs the tool with the right in hand and waits for it by reading the pipe
+// it is given until the tool closes it.
+//
+// AuthorizationExecuteWithPrivileges reports no exit status and does not say what
+// process it started, which is why the one-shot says so itself: what it prints is
+// the only evidence that the change was applied.
+func execute(ctx context.Context, authorization uintptr, self string, args []string) error {
+	var pinner runtime.Pinner
+	defer pinner.Unpin()
+
+	argv := make([]uintptr, 0, len(args)+1)
+	for _, arg := range args {
+		argv = append(argv, uintptr(unsafe.Pointer(cString(&pinner, arg))))
+	}
+	argv = append(argv, 0)
+	pinner.Pin(&argv[0])
+
+	var pipe uintptr
+	status := authorizationExecuteWithPrivileges(authorization, self, flagDefaults, &argv[0], &pipe)
+	switch status {
+	case errAuthorizationSuccess:
+	case errAuthorizationCanceled:
+		return ErrDeclined
+	case errAuthorizationToolExecuteFailure, errAuthorizationToolEnvironmentError:
+		// The right was granted and the tool still did not start. Nothing the user
+		// can do about it from here, so point them at the command instead.
+		return fmt.Errorf("%w: the system would not run %s elevated (OSStatus %d)", ErrUnavailable, self, status)
+	default:
+		return fmt.Errorf("run %s elevated: OSStatus %d", self, status)
+	}
+
+	out, err := readPipe(ctx, pipe)
+	if err != nil {
+		return err
+	}
+	return checkApplied(out)
+}
+
+// checkApplied reads the one-shot's report, which stands in for the exit status
+// there is no way to ask for here. A run that said nothing did not apply the
+// change, whatever else went on.
+func checkApplied(out string) error {
+	if !strings.Contains(out, AppliedMarker) {
+		return fmt.Errorf("elevated netbird did not report the change as applied: %s", firstLine(out))
+	}
+	return nil
+}
+
+// readPipe drains the tool's output, which ends when the tool exits and is
+// therefore also how we wait for it.
+func readPipe(ctx context.Context, pipe uintptr) (string, error) {
+	if pipe == 0 {
+		return "", nil
+	}
+	defer fclose(pipe)
+
+	fd := int(fileno(pipe))
+	if fd < 0 {
+		return "", nil
+	}
+
+	var out strings.Builder
+	buf := make([]byte, 4096)
+	for {
+		if err := ctx.Err(); err != nil {
+			return out.String(), err
+		}
+		n, err := syscall.Read(fd, buf)
+		if n > 0 {
+			out.Write(buf[:n])
+		}
+		switch {
+		case errors.Is(err, syscall.EINTR):
+			// A signal landed mid-read, which says nothing about the tool.
+			continue
+		case err != nil:
+			log.Debugf("read the elevated process's output: %v", err)
+			return out.String(), nil
+		case n <= 0:
+			// End of file: the tool closed the pipe, which is how it exiting
+			// reaches us.
+			return out.String(), nil
+		}
+	}
+}
+
+// itemSet builds an AuthorizationItemSet over items, pinned for the call.
+func itemSet(pinner *runtime.Pinner, items ...authorizationItem) *authorizationItemSet {
+	pinner.Pin(&items[0])
+	set := &authorizationItemSet{count: uint32(len(items)), items: &items[0]}
+	pinner.Pin(set)
+	return set
+}
+
+// promptItem is the environment entry carrying our sentence for the dialog.
+func promptItem(pinner *runtime.Pinner) authorizationItem {
+	value := []byte(promptText)
+	pinner.Pin(&value[0])
+	return authorizationItem{
+		name:        cString(pinner, promptKey),
+		valueLength: uintptr(len(value)),
+		value:       unsafe.Pointer(&value[0]),
+	}
+}
+
+// cString returns a NUL-terminated copy of s, pinned so the C side may hold it for
+// the duration of the call.
+func cString(pinner *runtime.Pinner, s string) *byte {
+	b := append([]byte(s), 0)
+	pinner.Pin(&b[0])
+	return &b[0]
+}

client/internal/elevate/run_darwin_test.go

@@ -0,0 +1,111 @@
+package elevate
+
+import (
+	"errors"
+	"runtime"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// The framework has to load and the symbols have to resolve, or nothing else here
+// means anything.
+func TestSecurityFrameworkLoads(t *testing.T) {
+	require.NoError(t, load(), "Security.framework must open")
+
+	for name, fn := range map[string]any{
+		"AuthorizationCreate":                authorizationCreate,
+		"AuthorizationExecuteWithPrivileges": authorizationExecuteWithPrivileges,
+		"AuthorizationFree":                  authorizationFree,
+		"fileno":                             fileno,
+		"fclose":                             fclose,
+	} {
+		assert.NotNil(t, fn, "%s must resolve", name)
+	}
+}
+
+// A request with no interaction allowed exercises the whole call — the rights and
+// environment structs, and the OSStatus that comes back — without a dialog anybody
+// has to answer. What the system decides is its business; that it decides at all is
+// what this asserts.
+func TestAuthorizationCreateWithoutInteraction(t *testing.T) {
+	if err := load(); err != nil {
+		t.Skipf("Security.framework did not open: %v", err)
+	}
+
+	var pinner runtime.Pinner
+	defer pinner.Unpin()
+
+	rights := itemSet(&pinner, authorizationItem{name: cString(&pinner, rightExecute)})
+	environment := itemSet(&pinner, promptItem(&pinner))
+	require.EqualValues(t, 1, rights.count, "the rights struct layout must match the C one")
+
+	var authorization uintptr
+	status := authorizationCreate(rights, environment, flagDefaults|flagExtendRights, &authorization)
+
+	switch status {
+	case errAuthorizationSuccess:
+		// Credentials were already cached for this session.
+		authorizationFree(authorization, flagDestroyRights)
+	case errAuthorizationDenied, errAuthorizationInteractionNotAllowed:
+		// The expected answers when nobody may be asked.
+	default:
+		require.Failf(t, "unknown OSStatus", "AuthorizationCreate returned %d, want a status we recognise", status)
+	}
+}
+
+// Asking with a right nobody has must not be mistaken for a declined prompt: the
+// caller would report nothing at all.
+func TestAuthorizeUnknownRightIsNotDeclined(t *testing.T) {
+	if err := load(); err != nil {
+		t.Skipf("Security.framework did not open: %v", err)
+	}
+
+	var pinner runtime.Pinner
+	defer pinner.Unpin()
+
+	rights := itemSet(&pinner, authorizationItem{name: cString(&pinner, "io.netbird.right.that.does.not.exist")})
+
+	var authorization uintptr
+	status := authorizationCreate(rights, nil, flagDefaults|flagExtendRights, &authorization)
+	if status == errAuthorizationSuccess {
+		authorizationFree(authorization, flagDestroyRights)
+	}
+	assert.NotEqual(t, int32(errAuthorizationSuccess), status, "a right that does not exist must not be granted")
+}
+
+func TestMechanismAvailable(t *testing.T) {
+	assert.True(t, mechanismAvailable(), "the trampoline exists on every macOS")
+}
+
+// The one-shot's report is what stands in for an exit status here, so a run that
+// says nothing must not read as success.
+func TestCheckApplied(t *testing.T) {
+	require.NoError(t, checkApplied(AppliedMarker+"\n"), "the report the one-shot prints")
+	require.NoError(t, checkApplied("some warning\n"+AppliedMarker+"\n"), "the report after other output")
+
+	assert.Error(t, checkApplied(""), "a run that printed nothing did not apply the change")
+	assert.Error(t, checkApplied("dyld: library not loaded\n"), "output that is not the report")
+}
+
+// A panic out of the FFI layer has to reach the caller as "no mechanism", which is
+// the outcome that offers the user the command instead of taking the window down.
+func TestGuardTurnsAPanicIntoUnavailable(t *testing.T) {
+	err := guard("pretending to call something", func() error {
+		panic("purego: signature it cannot map")
+	})
+
+	require.ErrorIs(t, err, ErrUnavailable, "a panic must read as a missing mechanism")
+	assert.Contains(t, err.Error(), "pretending to call something", "what panicked")
+}
+
+// guard wraps every darwin path, so what a caller switches on has to survive it.
+func TestGuardPassesErrorsThrough(t *testing.T) {
+	sentinel := errors.New("the call itself failed")
+	assert.ErrorIs(t, guard("calling", func() error { return sentinel }), sentinel,
+		"the error it was given")
+	assert.ErrorIs(t, guard("calling", func() error { return ErrDeclined }), ErrDeclined,
+		"a declined prompt stays declined")
+	assert.NoError(t, guard("calling", func() error { return nil }), "a call that worked")
+}

client/internal/elevate/run_unix.go

@@ -0,0 +1,117 @@
+//go:build linux || freebsd
+
+package elevate
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"io"
+	"os"
+	"os/exec"
+	"strings"
+)
+
+// pkexec exit codes that are about the authorization rather than about the program
+// we asked it to run. The manual page reserves both.
+const (
+	// exitDismissed is returned when the user dismissed the authentication
+	// dialog.
+	exitDismissed = 126
+	// exitNotAuthorized is returned when the authorization was not obtained. That
+	// covers the user saying no as well as pkexec having had nobody to ask: see
+	// noAgentMarkers.
+	exitNotAuthorized = 127
+)
+
+// exitNotAuthorized covers three different endings that only pkexec's own words
+// tell apart, so they are matched here. Read with LC_ALL=C so the words are the
+// ones written below.
+//
+// refusedMarker is a refusal: the user said no, gave up on the password, or holds
+// an account that may not elevate at all.
+const refusedMarker = "Not authorized"
+
+// noAgentMarkers say pkexec had no way to ask: no agent registered for the
+// session, and no controlling terminal for the textual agent it falls back to.
+var noAgentMarkers = []string{"authentication agent", "controlling terminal"}
+
+// run asks polkit to run self as root. pkexec hands the request to the session's
+// polkit agent, which is what prompts and what collects any password; we see only
+// its verdict.
+//
+// The environment is otherwise deliberately not passed through: pkexec clears it
+// bar a small allowlist, and the one-shot needs nothing from it.
+func run(ctx context.Context, self string, args []string) error {
+	pkexec, err := exec.LookPath("pkexec")
+	if err != nil {
+		return fmt.Errorf("%w: pkexec is not installed", ErrUnavailable)
+	}
+
+	cmd := exec.CommandContext(ctx, pkexec, append([]string{self}, args...)...)
+	// C locale so pkexec's own diagnostics are the ones noAgentMarkers knows.
+	cmd.Env = append(os.Environ(), "LC_ALL=C")
+	var stderr strings.Builder
+	cmd.Stderr = &stderr
+	// The one-shot reports itself on stdout for macOS's sake, where there is no
+	// exit status to read. Here there is one, so that line is noise.
+	cmd.Stdout = io.Discard
+
+	err = cmd.Run()
+	if err == nil {
+		return nil
+	}
+
+	var exitErr *exec.ExitError
+	if !errors.As(err, &exitErr) {
+		return fmt.Errorf("run pkexec: %w", err)
+	}
+
+	// Matched against everything pkexec said, reported as one line: a complaint
+	// that is not the first thing printed still has to be recognised, and reading
+	// it as a refusal would swallow it.
+	full := stderr.String()
+	out := firstLine(full)
+
+	switch exitErr.ExitCode() {
+	case exitDismissed:
+		return ErrDeclined
+	case exitNotAuthorized:
+		return notAuthorized(full, out)
+	default:
+		return fmt.Errorf("elevated netbird exited with %d: %s", exitErr.ExitCode(), out)
+	}
+}
+
+// notAuthorized sorts out the three endings pkexec reports as exitNotAuthorized.
+//
+// It also returns that code when the authorization succeeded and it then could
+// not run the program, so a refusal has to be recognised rather than assumed:
+// reading every one of these as "the user said no" would revert the control in
+// silence on a host where elevation is broken.
+func notAuthorized(full, out string) error {
+	switch {
+	case hasAny(full, noAgentMarkers):
+		return fmt.Errorf("%w: polkit had no way to ask: %s", ErrUnavailable, out)
+	case out == noOutput, strings.Contains(full, refusedMarker):
+		// The user said no, which needs no message; that an account barred from
+		// elevating altogether lands here too is why the reason is kept.
+		return fmt.Errorf("%w: %s", ErrDeclined, out)
+	default:
+		return fmt.Errorf("pkexec could not run elevated netbird: %s", out)
+	}
+}
+
+func hasAny(s string, markers []string) bool {
+	for _, marker := range markers {
+		if strings.Contains(s, marker) {
+			return true
+		}
+	}
+	return false
+}
+
+func mechanismAvailable() bool {
+	_, err := exec.LookPath("pkexec")
+	return err == nil
+}

client/internal/elevate/run_unix_test.go

@@ -0,0 +1,110 @@
+//go:build linux || freebsd
+
+package elevate
+
+import (
+	"context"
+	"fmt"
+	"os"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// fakePkexec puts a pkexec on PATH that exits with the given code, so the
+// mapping from polkit's exit codes onto our errors can be exercised without a
+// polkit agent.
+func fakePkexec(t *testing.T, exitCode int, stderr string) {
+	t.Helper()
+
+	dir := t.TempDir()
+	script := fmt.Sprintf("#!/bin/sh\necho %s >&2\nexit %d\n", shellQuote(stderr), exitCode)
+	require.NoError(t, os.WriteFile(filepath.Join(dir, "pkexec"), []byte(script), 0o700), "write the fake pkexec")
+	t.Setenv("PATH", dir)
+}
+
+func shellQuote(s string) string {
+	return "'" + strings.ReplaceAll(s, "'", `'\''`) + "'"
+}
+
+func TestRunMapsPkexecExitCodes(t *testing.T) {
+	tests := []struct {
+		name     string
+		exitCode int
+		stderr   string
+		wantErr  error
+	}{
+		{name: "applied", exitCode: 0},
+		{
+			name:     "dialog dismissed",
+			exitCode: exitDismissed,
+			stderr:   "Error executing command as another user: Request dismissed",
+			wantErr:  ErrDeclined,
+		},
+		{
+			// What a graphical agent reports for a cancelled prompt. Not a
+			// failure: the user was asked and answered.
+			name:     "prompt cancelled",
+			exitCode: exitNotAuthorized,
+			stderr:   "Error executing command as another user: Not authorized",
+			wantErr:  ErrDeclined,
+		},
+		{
+			// The same status, but pkexec never got to ask anybody.
+			name:     "no agent and no terminal to fall back on",
+			exitCode: exitNotAuthorized,
+			stderr:   "Error creating textual authentication agent: Error opening current controlling terminal for the process (`/dev/tty'): No such device or address",
+			wantErr:  ErrUnavailable,
+		},
+		{
+			// And the same status again once the authorization succeeded and
+			// pkexec could not run what it had been authorized to run. Reading
+			// that as a refusal would revert the control in silence on a host
+			// where elevation is broken.
+			name:     "authorized but not runnable",
+			exitCode: exitNotAuthorized,
+			stderr:   "Error executing command as another user: No such file or directory",
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			fakePkexec(t, tt.exitCode, tt.stderr)
+
+			err := run(context.Background(), "/nonexistent/netbird-ui", []string{"--flag"})
+			switch {
+			case tt.wantErr != nil:
+				require.ErrorIs(t, err, tt.wantErr, "exit %d said %q", tt.exitCode, tt.stderr)
+			case tt.exitCode == 0:
+				require.NoError(t, err, "a pkexec that exited cleanly applied the change")
+			default:
+				require.Error(t, err, "exit %d said %q", tt.exitCode, tt.stderr)
+				assert.NotErrorIs(t, err, ErrDeclined, "not the user's answer")
+				assert.NotErrorIs(t, err, ErrUnavailable, "not a missing mechanism")
+			}
+		})
+	}
+}
+
+// An exit code that is not polkit's is the one-shot's own failure, and has to
+// stay distinguishable from a declined prompt: the caller reports it.
+func TestRunReportsOneShotFailure(t *testing.T) {
+	fakePkexec(t, 3, "the one-shot said no")
+
+	err := run(context.Background(), "/nonexistent/netbird-ui", nil)
+
+	require.Error(t, err, "a one-shot that failed is not a prompt that was answered")
+	assert.NotErrorIs(t, err, ErrDeclined, "not the user's answer")
+	assert.NotErrorIs(t, err, ErrUnavailable, "not a missing mechanism")
+}
+
+func TestRunWithoutPkexecIsUnavailable(t *testing.T) {
+	t.Setenv("PATH", t.TempDir())
+
+	err := run(context.Background(), "/nonexistent/netbird-ui", nil)
+	require.ErrorIs(t, err, ErrUnavailable, "no pkexec means no mechanism")
+	assert.False(t, mechanismAvailable(), "mechanismAvailable without pkexec on PATH")
+}

client/internal/elevate/run_unsupported.go

@@ -0,0 +1,15 @@
+//go:build !windows && !darwin && !linux && !freebsd
+
+package elevate
+
+import "context"
+
+// run reports that this platform has no elevation prompt to drive. Mobile and
+// WASM builds have no local user to ask in the first place.
+func run(context.Context, string, []string) error {
+	return ErrUnavailable
+}
+
+func mechanismAvailable() bool {
+	return false
+}

client/internal/elevate/run_windows.go

@@ -0,0 +1,193 @@
+package elevate
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"runtime"
+	"unsafe"
+
+	log "github.com/sirupsen/logrus"
+	"golang.org/x/sys/windows"
+)
+
+const (
+	// seeMaskNoCloseProcess keeps the started process's handle open in
+	// hProcess so we can wait for it.
+	seeMaskNoCloseProcess = 0x00000040
+	// seeMaskNoAsync makes ShellExecuteExW finish its work before returning,
+	// which it must when the calling thread does not pump messages.
+	seeMaskNoAsync = 0x00000100
+	// seeMaskFlagNoUI suppresses the shell's own error dialogs; the UAC consent
+	// dialog is not one of them and still appears.
+	seeMaskFlagNoUI = 0x00000400
+
+	// swHide: the one-shot has no window to show.
+	swHide = 0
+
+	// sFalse (S_FALSE) answers CoInitializeEx when COM is already up on this
+	// thread in the mode we asked for; rpcChangedMode (RPC_E_CHANGED_MODE) when
+	// it is up in the other one.
+	sFalse         = 1
+	rpcChangedMode = 0x80010106
+)
+
+// shellExecuteInfoW mirrors SHELLEXECUTEINFOW. The field order and Go's own
+// padding match the C layout on both 386 and amd64.
+type shellExecuteInfoW struct {
+	cbSize         uint32
+	fMask          uint32
+	hwnd           windows.HWND
+	lpVerb         *uint16
+	lpFile         *uint16
+	lpParameters   *uint16
+	lpDirectory    *uint16
+	nShow          int32
+	hInstApp       windows.Handle
+	lpIDList       uintptr
+	lpClass        *uint16
+	hkeyClass      windows.Handle
+	dwHotKey       uint32
+	hIconOrMonitor windows.Handle
+	hProcess       windows.Handle
+}
+
+var (
+	shell32            = windows.NewLazySystemDLL("shell32.dll")
+	procShellExecuteEx = shell32.NewProc("ShellExecuteExW")
+)
+
+// run starts self elevated with the "runas" verb, which is what raises the UAC
+// consent dialog, and waits for it to finish. Windows decides whether consent is
+// enough or an administrator's credentials are needed, and collects them itself.
+func run(ctx context.Context, self string, args []string) error {
+	verb, err := windows.UTF16PtrFromString("runas")
+	if err != nil {
+		return fmt.Errorf("encode verb: %w", err)
+	}
+	file, err := windows.UTF16PtrFromString(self)
+	if err != nil {
+		return fmt.Errorf("encode %s: %w", self, err)
+	}
+	params, err := windows.UTF16PtrFromString(windows.ComposeCommandLine(args))
+	if err != nil {
+		return fmt.Errorf("encode arguments: %w", err)
+	}
+
+	info := shellExecuteInfoW{
+		fMask:        seeMaskNoCloseProcess | seeMaskNoAsync | seeMaskFlagNoUI,
+		hwnd:         ownerWindow(),
+		lpVerb:       verb,
+		lpFile:       file,
+		lpParameters: params,
+		nShow:        swHide,
+	}
+	info.cbSize = uint32(unsafe.Sizeof(info))
+
+	process, err := shellExecute(&info)
+	if err != nil {
+		return err
+	}
+	defer func() {
+		if err := windows.CloseHandle(process); err != nil {
+			log.Debugf("close elevated process handle: %v", err)
+		}
+	}()
+
+	return waitForProcess(ctx, process)
+}
+
+// shellExecute performs the call itself. ShellExecuteExW wants COM initialised on
+// the calling thread, so the goroutine is pinned to one for the duration and COM
+// is set up on it; an "already initialised, different mode" answer is fine,
+// because then somebody else has done it for us.
+func shellExecute(info *shellExecuteInfoW) (windows.Handle, error) {
+	runtime.LockOSThread()
+	defer runtime.UnlockOSThread()
+
+	switch err := windows.CoInitializeEx(0, windows.COINIT_APARTMENTTHREADED); {
+	case err == nil, isHResult(err, sFalse):
+		// Ours, or already initialised in the same mode: either way this call
+		// counts and has to be balanced.
+		defer windows.CoUninitialize()
+	case isHResult(err, rpcChangedMode):
+		// The thread is already in the other apartment model. ShellExecuteExW
+		// works there too, and there is nothing of ours to balance.
+	default:
+		return 0, fmt.Errorf("initialise COM: %w", err)
+	}
+
+	ret, _, lastErr := procShellExecuteEx.Call(uintptr(unsafe.Pointer(info)))
+	if ret != 0 {
+		return info.hProcess, nil
+	}
+
+	if errors.Is(lastErr, windows.ERROR_CANCELLED) {
+		return 0, ErrDeclined
+	}
+	return 0, fmt.Errorf("run elevated: %w", lastErr)
+}
+
+// ownerWindow returns this process's foreground window, and 0 when the window in
+// front belongs to somebody else or cannot be attributed. ShellExecuteExW takes it
+// as the parent for the UI it raises, which is what keeps the consent dialog in
+// front of the window the user was just clicking in instead of behind it. It is
+// also what a remote-desktop session needs to place the dialog at all when the
+// secure desktop is switched off.
+func ownerWindow() windows.HWND {
+	hwnd := windows.GetForegroundWindow()
+	if hwnd == 0 {
+		return 0
+	}
+
+	var pid uint32
+	if _, err := windows.GetWindowThreadProcessId(hwnd, &pid); err != nil {
+		log.Debugf("cannot attribute the foreground window, raising the prompt without an owner: %v", err)
+		return 0
+	}
+	if pid != windows.GetCurrentProcessId() {
+		return 0
+	}
+	return hwnd
+}
+
+// isHResult reports whether err carries the given HRESULT. CoInitializeEx
+// returns its HRESULT as an Errno, so the comparison is on the raw value.
+func isHResult(err error, hresult uintptr) bool {
+	var errno windows.Errno
+	return errors.As(err, &errno) && uintptr(errno) == hresult
+}
+
+func waitForProcess(ctx context.Context, process windows.Handle) error {
+	// The wait is interruptible so a cancelled context stops us waiting on a
+	// consent dialog nobody is going to answer. The elevated process is not
+	// ours to kill, and it either applies the change or does not.
+	for {
+		event, err := windows.WaitForSingleObject(process, 250)
+		if err != nil {
+			return fmt.Errorf("wait for the elevated process: %w", err)
+		}
+		if event == uint32(windows.WAIT_OBJECT_0) {
+			break
+		}
+		if err := ctx.Err(); err != nil {
+			return err
+		}
+	}
+
+	var code uint32
+	if err := windows.GetExitCodeProcess(process, &code); err != nil {
+		return fmt.Errorf("read the elevated process's exit code: %w", err)
+	}
+	if code != 0 {
+		return fmt.Errorf("elevated netbird exited with %d", code)
+	}
+	return nil
+}
+
+// mechanismAvailable is true on Windows: UAC prompts for consent when the user
+// is an administrator and for an administrator's credentials when they are not,
+// so there is always something to ask.
+func mechanismAvailable() bool {
+	return true
+}

client/internal/elevate/trusted.go

@@ -0,0 +1,40 @@
+package elevate
+
+import (
+	"fmt"
+	"os"
+	"path/filepath"
+)
+
+// trustedSelf returns the path of this executable, provided it is one we are
+// willing to have run as root.
+//
+// The check is what keeps elevation from becoming a way to launder someone
+// else's code into a root process: the user consents to NetBird being elevated,
+// having been shown NetBird's name, so what runs must be the file NetBird was
+// installed as and not something a third party could have swapped for it. An
+// executable only its owner can write is that; anything wider is refused, and
+// the caller falls back to showing the command instead.
+//
+// The owner writing to their own executable is not part of that threat: code
+// running as the user can already prompt them for anything, and could just as
+// well ask them to run the command by hand. What matters is that no *other*
+// unprivileged account can reach it.
+func trustedSelf() (string, error) {
+	exe, err := os.Executable()
+	if err != nil {
+		return "", fmt.Errorf("locate this executable: %w", err)
+	}
+
+	// Resolve symlinks so the checks below apply to the file that would actually
+	// be executed, not to a link somebody else may control.
+	resolved, err := filepath.EvalSymlinks(exe)
+	if err != nil {
+		return "", fmt.Errorf("resolve %s: %w", exe, err)
+	}
+
+	if err := checkOnlyOwnerWritable(resolved); err != nil {
+		return "", fmt.Errorf("%w: %s cannot be trusted to run as root: %w", ErrUnavailable, resolved, err)
+	}
+	return resolved, nil
+}

client/internal/elevate/trusted_group_darwin.go

@@ -0,0 +1,10 @@
+package elevate
+
+// adminWriteGIDs are the groups whose write access to an executable does not
+// widen who could authorize elevating it.
+//
+// macOS installs applications as root:admin, mode 0775, /Applications included,
+// so requiring owner-only write would reject every normal install. Group admin
+// (gid 80) is exactly the set of accounts that can answer the authentication
+// dialog, so its write access grants nothing the prompt would not.
+var adminWriteGIDs = []uint32{0, 80}

client/internal/elevate/trusted_group_unix.go

@@ -0,0 +1,9 @@
+//go:build !windows && !darwin
+
+package elevate
+
+// adminWriteGIDs are the groups whose write access to an executable does not
+// widen who could authorize elevating it. Only root's own group qualifies here:
+// a distribution installs into root-owned directories, and there is no
+// system-wide administrators group that both writes them and answers polkit.
+var adminWriteGIDs = []uint32{0}

client/internal/elevate/trusted_unix.go

@@ -0,0 +1,146 @@
+//go:build !windows
+
+package elevate
+
+import (
+	"bufio"
+	"errors"
+	"fmt"
+	"os"
+	"os/user"
+	"path/filepath"
+	"slices"
+	"strconv"
+	"strings"
+	"syscall"
+
+	log "github.com/sirupsen/logrus"
+)
+
+// groupFile lists which accounts are in which group, for the membership a user
+// private group's name does not state: see groupHasOtherMembers.
+const groupFile = "/etc/group"
+
+// checkOnlyOwnerWritable reports an error unless path, and every directory leading
+// to it, is owned by either root or this user and writable by nobody who could not
+// already act as its owner. A writable directory is as good as a writable file,
+// since anything in it can be replaced, so the whole chain is checked.
+func checkOnlyOwnerWritable(path string) error {
+	self := uint32(os.Getuid())
+
+	for dir := path; ; dir = filepath.Dir(dir) {
+		info, err := os.Lstat(dir)
+		if err != nil {
+			return fmt.Errorf("stat %s: %w", dir, err)
+		}
+
+		stat, ok := info.Sys().(*syscall.Stat_t)
+		if !ok {
+			return errors.New("file ownership is unavailable on this platform")
+		}
+		if stat.Uid != 0 && stat.Uid != self {
+			return fmt.Errorf("%s is owned by uid %d, neither root nor this user", dir, stat.Uid)
+		}
+
+		if err := checkWriteBits(dir, info, stat.Uid, stat.Gid); err != nil {
+			return err
+		}
+
+		if parent := filepath.Dir(dir); parent == dir {
+			return nil
+		}
+	}
+}
+
+func checkWriteBits(path string, info os.FileInfo, uid, gid uint32) error {
+	// On a directory the sticky bit stands in for the write bits: whoever may
+	// write there still cannot replace an entry they do not own, which is the
+	// only thing that would matter to us. /tmp is the usual example.
+	sticky := info.IsDir() && info.Mode()&os.ModeSticky != 0
+
+	return writeBitsAllow(path, info.Mode().Perm(), sticky, groupWriteAllowed(uid, gid))
+}
+
+// writeBitsAllow decides on the permission bits alone, given whether the group's
+// write access has been vouched for.
+func writeBitsAllow(path string, perm os.FileMode, sticky, groupAllowed bool) error {
+	if sticky {
+		return nil
+	}
+	if perm&0o020 != 0 && !groupAllowed {
+		return fmt.Errorf("%s is writable by a group with members other than its owner (%v)", path, perm)
+	}
+	if perm&0o002 != 0 {
+		return fmt.Errorf("%s is world-writable (%v)", path, perm)
+	}
+	return nil
+}
+
+// groupWriteAllowed reports whether a group's write access to a file owned by uid
+// puts it in reach of anyone who could not already act as that owner.
+//
+// Two ways it does not. A group in adminWriteGIDs holds the accounts that can
+// answer the elevation prompt anyway. And a user private group is how Debian,
+// Ubuntu and Fedora ship: their umask of 002 makes a home directory and
+// everything built in it group-writable, so refusing that would refuse every
+// build not installed from a package.
+func groupWriteAllowed(uid, gid uint32) bool {
+	if slices.Contains(adminWriteGIDs, gid) {
+		return true
+	}
+
+	group, err := user.LookupGroupId(strconv.FormatUint(uint64(gid), 10))
+	if err != nil {
+		log.Debugf("cannot look up group %d, treating it as shared: %v", gid, err)
+		return false
+	}
+	owner, err := user.LookupId(strconv.FormatUint(uint64(uid), 10))
+	if err != nil {
+		log.Debugf("cannot look up uid %d, treating its group as shared: %v", uid, err)
+		return false
+	}
+
+	if group.Name != owner.Username {
+		return false
+	}
+	return !groupHasOtherMembers(groupFile, group.Name, owner.Username)
+}
+
+// groupHasOtherMembers reports whether the group lists a member besides owner.
+//
+// Sharing the owner's name is what a user private group is recognised by, and it
+// says nothing about who is in it: a group that has since gained a member is
+// still named that way, and that member can write whatever the group can. So the
+// membership is read rather than assumed. A group this file does not describe,
+// because it comes from LDAP or another NSS source, cannot be answered here and
+// leaves the name as the only thing to go on.
+func groupHasOtherMembers(path, name, owner string) bool {
+	file, err := os.Open(path)
+	if err != nil {
+		log.Debugf("cannot read %s for the members of group %q: %v", path, name, err)
+		return false
+	}
+	defer func() {
+		if err := file.Close(); err != nil {
+			log.Debugf("close %s: %v", path, err)
+		}
+	}()
+
+	scanner := bufio.NewScanner(file)
+	for scanner.Scan() {
+		// name:password:gid:member,member
+		fields := strings.Split(scanner.Text(), ":")
+		if len(fields) < 4 || fields[0] != name {
+			continue
+		}
+		for member := range strings.SplitSeq(fields[3], ",") {
+			if member != "" && member != owner {
+				return true
+			}
+		}
+	}
+	if err := scanner.Err(); err != nil {
+		log.Debugf("read %s: %v", path, err)
+	}
+	return false
+}

client/internal/elevate/trusted_unix_test.go

@@ -0,0 +1,177 @@
+//go:build !windows
+
+package elevate
+
+import (
+	"os"
+	"os/user"
+	"path/filepath"
+	"strconv"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// ownerOnlyDir is t.TempDir() with the write bits tightened. testing creates its
+// numbered directory with 0777 minus the umask, so under the common 002 umask it
+// is group-writable and would fail the check under test on its own.
+func ownerOnlyDir(t *testing.T) string {
+	t.Helper()
+	dir := t.TempDir()
+	require.NoError(t, os.Chmod(dir, 0o755), "tighten the temporary directory")
+	return dir
+}
+
+// writeExecutable creates a plain executable file, the shape trustedSelf checks.
+func writeExecutable(t *testing.T, dir string) string {
+	t.Helper()
+	path := filepath.Join(dir, "netbird-ui")
+	require.NoError(t, os.WriteFile(path, []byte("#!/bin/sh\n"), 0o755), "write the executable")
+	require.NoError(t, os.Chmod(path, 0o755), "set the executable's mode")
+	return path
+}
+
+func TestCheckOnlyOwnerWritableAcceptsOwnerOnly(t *testing.T) {
+	err := checkOnlyOwnerWritable(writeExecutable(t, ownerOnlyDir(t)))
+	assert.NoError(t, err, "an owner-only writable executable is trustworthy")
+}
+
+func TestCheckOnlyOwnerWritableRejectsWorldWritableFile(t *testing.T) {
+	path := writeExecutable(t, ownerOnlyDir(t))
+	require.NoError(t, os.Chmod(path, 0o777), "make the executable world-writable")
+
+	assert.Error(t, checkOnlyOwnerWritable(path), "a world-writable executable must be refused")
+}
+
+// The permission policy on its own, without a filesystem to arrange: whether the
+// group has been vouched for is the only thing that makes group write acceptable.
+func TestWriteBitsAllow(t *testing.T) {
+	tests := []struct {
+		name         string
+		perm         os.FileMode
+		sticky       bool
+		groupAllowed bool
+		wantErr      bool
+	}{
+		{name: "owner only", perm: 0o755},
+		{name: "group write in a private group", perm: 0o775, groupAllowed: true},
+		{name: "group write in a shared group", perm: 0o775, wantErr: true},
+		{name: "world write", perm: 0o777, groupAllowed: true, wantErr: true},
+		{name: "world write on a sticky directory", perm: 0o777, sticky: true},
+		{name: "group write on a sticky directory", perm: 0o775, sticky: true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			err := writeBitsAllow("/path", tt.perm, tt.sticky, tt.groupAllowed)
+			if tt.wantErr {
+				assert.Error(t, err, "perm %v, sticky %v, group allowed %v", tt.perm, tt.sticky, tt.groupAllowed)
+				return
+			}
+			assert.NoError(t, err, "perm %v, sticky %v, group allowed %v", tt.perm, tt.sticky, tt.groupAllowed)
+		})
+	}
+}
+
+// A build under a home directory on a distribution with a 002 umask, which is what
+// a locally built or tarball-installed binary looks like. Its group has no members
+// but its owner, so it is as good as owner-only.
+//
+// Whether this host is such a distribution is read from the environment rather than
+// from groupWriteAllowed: asking the function under test whether to run would let
+// it skip its own coverage away if it regressed to refusing everything.
+func TestCheckOnlyOwnerWritableAcceptsOwnPrivateGroup(t *testing.T) {
+	requirePrivatePrimaryGroup(t)
+
+	dir := ownerOnlyDir(t)
+	path := writeExecutable(t, dir)
+	require.NoError(t, os.Chmod(dir, 0o775), "make the directory group-writable")
+	require.NoError(t, os.Chmod(path, 0o775), "make the executable group-writable")
+
+	err := checkOnlyOwnerWritable(path)
+	assert.NoError(t, err, "group write in the owner's own private group reaches nobody else")
+}
+
+// A group that shares its owner's name but has gained another member is no longer
+// private, and its write access reaches an account that could not elevate.
+func TestGroupHasOtherMembers(t *testing.T) {
+	tests := []struct {
+		name  string
+		entry string
+		want  bool
+	}{
+		{name: "no members", entry: "vma:x:1000:"},
+		{name: "only the owner", entry: "vma:x:1000:vma"},
+		{name: "another member", entry: "vma:x:1000:bob", want: true},
+		{name: "the owner and another", entry: "vma:x:1000:vma,bob", want: true},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			path := filepath.Join(t.TempDir(), "group")
+			body := "root:x:0:\n" + tt.entry + "\nsudo:x:27:vma\n"
+			require.NoError(t, os.WriteFile(path, []byte(body), 0o644), "write the group file")
+
+			assert.Equal(t, tt.want, groupHasOtherMembers(path, "vma", "vma"), "entry %q", tt.entry)
+		})
+	}
+}
+
+// A group file that says nothing about the group leaves the name as the only thing
+// to go on, so the private-group allowance stands rather than collapsing on every
+// host whose groups come from LDAP.
+func TestGroupHasOtherMembersTolerantOfAnUnknownGroup(t *testing.T) {
+	path := filepath.Join(t.TempDir(), "group")
+	require.NoError(t, os.WriteFile(path, []byte("root:x:0:\n"), 0o644), "write the group file")
+
+	assert.False(t, groupHasOtherMembers(path, "vma", "vma"), "a group the file does not describe")
+	assert.False(t, groupHasOtherMembers(filepath.Join(t.TempDir(), "absent"), "vma", "vma"),
+		"no group file at all")
+}
+
+// A writable directory is as good as a writable file: whoever can write the
+// directory can put a different binary at the same path.
+func TestCheckOnlyOwnerWritableRejectsWritableDirectory(t *testing.T) {
+	dir := filepath.Join(ownerOnlyDir(t), "bin")
+	require.NoError(t, os.Mkdir(dir, 0o755), "create the directory")
+	path := writeExecutable(t, dir)
+	require.NoError(t, os.Chmod(dir, 0o777), "make the directory world-writable")
+
+	assert.Error(t, checkOnlyOwnerWritable(path), "an executable in a world-writable directory must be refused")
+}
+
+// A sticky world-writable directory is exempt: the sticky bit is what stops one
+// user replacing another's entries. /tmp is why this matters.
+func TestCheckOnlyOwnerWritableAcceptsStickyDirectory(t *testing.T) {
+	dir := filepath.Join(ownerOnlyDir(t), "sticky")
+	require.NoError(t, os.Mkdir(dir, 0o755), "create the directory")
+	path := writeExecutable(t, dir)
+	require.NoError(t, os.Chmod(dir, 0o777|os.ModeSticky), "make the directory sticky and world-writable")
+
+	err := checkOnlyOwnerWritable(path)
+	assert.NoError(t, err, "the sticky bit stops another user replacing the executable")
+}
+
+func TestCheckOnlyOwnerWritableRejectsMissingFile(t *testing.T) {
+	err := checkOnlyOwnerWritable(filepath.Join(ownerOnlyDir(t), "absent"))
+	assert.Error(t, err, "an executable that is not there must be refused")
+}
+
+// requirePrivatePrimaryGroup skips unless this user's primary group is their own,
+// which is what the user-private-group allowance is about.
+func requirePrivatePrimaryGroup(t *testing.T) {
+	t.Helper()
+
+	self, err := user.Current()
+	require.NoError(t, err, "look up the test user")
+	group, err := user.LookupGroupId(strconv.Itoa(os.Getgid()))
+	require.NoError(t, err, "look up the test user's primary group")
+
+	if group.Name != self.Username {
+		t.Skipf("the test user's primary group is %q, not their own, so there is nothing to assert here", group.Name)
+	}
+	if groupHasOtherMembers(groupFile, group.Name, self.Username) {
+		t.Skipf("group %q has other members, so it is not a private group", group.Name)
+	}
+}

client/internal/elevate/trusted_windows.go

@@ -0,0 +1,215 @@
+package elevate
+
+import (
+	"errors"
+	"fmt"
+	"path/filepath"
+	"slices"
+	"unsafe"
+
+	"golang.org/x/sys/windows"
+)
+
+const (
+	// fileDeleteChild is FILE_DELETE_CHILD, which x/sys does not define: the
+	// right to delete an entry of a directory without holding DELETE on it.
+	fileDeleteChild = 0x00000040
+
+	// accessAllowedCallbackACEType is an allow ACE with a condition appended to
+	// the ACCESS_ALLOWED_ACE layout, so its trustee is still at SidStart.
+	accessAllowedCallbackACEType = 0x9
+
+	// The allow ACE types that carry object GUIDs ahead of the trustee, so the
+	// SID is not at SidStart. They occur on directory-service objects rather
+	// than files, and are refused rather than skipped: see aceTrustee.
+	accessAllowedObjectACEType         = 0x5
+	accessAllowedCallbackObjectACEType = 0xB
+)
+
+// fileWriteAccess are the rights that let a trustee rewrite or replace a file,
+// or take it over and then do so.
+const fileWriteAccess = windows.FILE_WRITE_DATA | windows.FILE_APPEND_DATA |
+	windows.DELETE | windows.WRITE_DAC | windows.WRITE_OWNER |
+	windows.GENERIC_WRITE | windows.GENERIC_ALL
+
+// dirWriteAccess are the rights over a directory that let a trustee replace an
+// entry somebody else owns. Creating a new entry is not one of them, which is
+// what the Unix sticky bit says in one bit: the root of every volume grants
+// BUILTIN\Users the right to add directories under it, and that reaches nothing
+// already there.
+const dirWriteAccess = fileDeleteChild | windows.DELETE |
+	windows.WRITE_DAC | windows.WRITE_OWNER | windows.GENERIC_ALL
+
+// trustedInstallerSID owns much of what Windows itself installs. x/sys has no
+// well-known constant for it.
+const trustedInstallerSID = "S-1-5-80-956008885-3418522649-1831038044-1853292631-2271478464"
+
+// checkOnlyOwnerWritable reports an error unless path, and every directory
+// leading to it, is owned by an account that can elevate (or by this user) and
+// grants write access to nobody else. A writable directory is as good as a
+// writable file, since an entry in it can be replaced, so the whole chain is
+// checked.
+func checkOnlyOwnerWritable(path string) error {
+	owners, err := trustedOwners()
+	if err != nil {
+		return err
+	}
+	writers, err := trustedWriters(owners)
+	if err != nil {
+		return err
+	}
+
+	writeAccess := windows.ACCESS_MASK(fileWriteAccess)
+	for target := path; ; target = filepath.Dir(target) {
+		if err := checkSecurity(target, writeAccess, owners, writers); err != nil {
+			return err
+		}
+		if parent := filepath.Dir(target); parent == target {
+			return nil
+		}
+		writeAccess = dirWriteAccess
+	}
+}
+
+// trustedOwners are the accounts we accept as the owner of the executable and of
+// the directories above it: the ones that can already answer the UAC prompt,
+// plus this user, whose own executable is theirs to write. Code running as the
+// user could prompt them for anything anyway; what matters is that no *other*
+// unprivileged account can reach it.
+func trustedOwners() ([]*windows.SID, error) {
+	self, err := currentUserSID()
+	if err != nil {
+		return nil, err
+	}
+
+	owners := []*windows.SID{self}
+	for _, wellKnown := range []windows.WELL_KNOWN_SID_TYPE{
+		windows.WinLocalSystemSid,
+		windows.WinBuiltinAdministratorsSid,
+	} {
+		sid, err := windows.CreateWellKnownSid(wellKnown)
+		if err != nil {
+			return nil, fmt.Errorf("build well-known SID %d: %w", wellKnown, err)
+		}
+		owners = append(owners, sid)
+	}
+
+	installer, err := windows.StringToSid(trustedInstallerSID)
+	if err != nil {
+		return nil, fmt.Errorf("parse TrustedInstaller SID: %w", err)
+	}
+	return append(owners, installer), nil
+}
+
+// trustedWriters are the trustees whose write access does not widen who could
+// decide what runs behind the prompt. The owners, and CREATOR OWNER, which
+// resolves to the object's owner and is therefore already vetted.
+func trustedWriters(owners []*windows.SID) ([]*windows.SID, error) {
+	creatorOwner, err := windows.CreateWellKnownSid(windows.WinCreatorOwnerSid)
+	if err != nil {
+		return nil, fmt.Errorf("build the CREATOR OWNER SID: %w", err)
+	}
+	return append(slices.Clone(owners), creatorOwner), nil
+}
+
+func checkSecurity(path string, writeAccess windows.ACCESS_MASK, owners, writers []*windows.SID) error {
+	sd, err := windows.GetNamedSecurityInfo(path, windows.SE_FILE_OBJECT,
+		windows.OWNER_SECURITY_INFORMATION|windows.DACL_SECURITY_INFORMATION)
+	if err != nil {
+		return fmt.Errorf("read security descriptor of %s: %w", path, err)
+	}
+
+	owner, _, err := sd.Owner()
+	if err != nil {
+		return fmt.Errorf("read owner of %s: %w", path, err)
+	}
+	if !containsSID(owners, owner) {
+		return fmt.Errorf("%s is owned by %s, which is neither this user nor an account that can elevate", path, owner)
+	}
+
+	dacl, _, err := sd.DACL()
+	if err != nil {
+		return fmt.Errorf("read DACL of %s: %w", path, err)
+	}
+	// A NULL DACL grants everyone everything; only an absent security
+	// descriptor would have got us here without one, and neither is trustworthy.
+	if dacl == nil {
+		return fmt.Errorf("%s has no DACL, so it grants write access to everyone", path)
+	}
+
+	return checkDACL(path, dacl, writeAccess, writers)
+}
+
+// checkDACL refuses an ACL that grants write access to a trustee outside
+// writers.
+//
+// An allowlist, because the trustees that must not have it cannot be listed: an
+// ACE naming an ordinary user account hands that account the same power as one
+// naming Everyone, and only the accounts that may hold it are knowable.
+func checkDACL(path string, dacl *windows.ACL, writeAccess windows.ACCESS_MASK, writers []*windows.SID) error {
+	for i := uint32(0); i < uint32(dacl.AceCount); i++ {
+		var ace *windows.ACCESS_ALLOWED_ACE
+		if err := windows.GetAce(dacl, i, &ace); err != nil {
+			return fmt.Errorf("read ACE %d of %s: %w", i, path, err)
+		}
+		// An inherit-only ACE says what children of this object get, not what
+		// this object grants.
+		if ace.Header.AceFlags&windows.INHERIT_ONLY_ACE != 0 {
+			continue
+		}
+		if ace.Mask&writeAccess == 0 {
+			continue
+		}
+		// Only an allow ACE grants anything; a deny ACE narrows what one gave.
+		if !isAllowACE(ace.Header.AceType) {
+			continue
+		}
+
+		trustee, err := aceTrustee(ace)
+		if err != nil {
+			return fmt.Errorf("read the trustee of ACE %d of %s: %w", i, path, err)
+		}
+		if !containsSID(writers, trustee) {
+			return fmt.Errorf("%s grants write access to %s", path, trustee)
+		}
+	}
+	return nil
+}
+
+// isAllowACE reports whether an ACE type grants rights, rather than denying,
+// auditing or labelling them.
+func isAllowACE(aceType uint8) bool {
+	switch aceType {
+	case windows.ACCESS_ALLOWED_ACE_TYPE, accessAllowedCallbackACEType,
+		accessAllowedObjectACEType, accessAllowedCallbackObjectACEType:
+		return true
+	default:
+		return false
+	}
+}
+
+// aceTrustee returns who an allow ACE grants its rights to. An ACE whose trustee
+// cannot be located is an error rather than something to skip past: being unable
+// to read who is being given write access is a refusal.
+func aceTrustee(ace *windows.ACCESS_ALLOWED_ACE) (*windows.SID, error) {
+	switch ace.Header.AceType {
+	case windows.ACCESS_ALLOWED_ACE_TYPE, accessAllowedCallbackACEType:
+		//nolint:gosec // SidStart is the first uint32 of the variable-length SID that follows the ACE header.
+		return (*windows.SID)(unsafe.Pointer(&ace.SidStart)), nil
+	default:
+		return nil, errors.New("an object-type allow ACE does not carry its trustee where we can read it")
+	}
+}
+
+func containsSID(sids []*windows.SID, sid *windows.SID) bool {
+	return slices.ContainsFunc(sids, sid.Equals)
+}
+
+func currentUserSID() (*windows.SID, error) {
+	token := windows.GetCurrentProcessToken()
+	user, err := token.GetTokenUser()
+	if err != nil {
+		return nil, fmt.Errorf("read this process's user: %w", err)
+	}
+	return user.User.Sid, nil
+}

client/internal/elevate/trusted_windows_test.go

@@ -0,0 +1,126 @@
+package elevate
+
+import (
+	"os"
+	"path/filepath"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"golang.org/x/sys/windows"
+)
+
+// A file the test user created under their own profile, which is what a per-user
+// install looks like. The whole chain up to the volume root is walked, so this is
+// also what says the walk does not refuse an ordinary Windows installation: the
+// root of every volume grants BUILTIN\Users rights that are not ours to worry
+// about.
+func TestCheckOnlyOwnerWritableAcceptsOwnFile(t *testing.T) {
+	err := checkOnlyOwnerWritable(writeExecutable(t))
+	assert.NoError(t, err, "a file the test user owns, under directories only administrators can write")
+}
+
+// Write access held by an account that cannot answer the UAC prompt means that
+// account decides what runs behind it, whoever the ACE names. The trustees that
+// must not have it cannot be listed, so the check names the ones that may.
+func TestCheckOnlyOwnerWritableRejectsUntrustedWriters(t *testing.T) {
+	tests := []struct {
+		name      string
+		wellKnown windows.WELL_KNOWN_SID_TYPE
+	}{
+		{name: "everyone", wellKnown: windows.WinWorldSid},
+		{name: "authenticated users", wellKnown: windows.WinAuthenticatedUserSid},
+		{name: "builtin users", wellKnown: windows.WinBuiltinUsersSid},
+		// A service account, which no denylist of the obvious groups would name
+		// and which cannot elevate any more than Everyone can.
+		{name: "local service", wellKnown: windows.WinLocalServiceSid},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			path := writeExecutable(t)
+			grantWrite(t, path, tt.wellKnown)
+
+			assert.Error(t, checkOnlyOwnerWritable(path),
+				"write access for %s must be refused", tt.name)
+		})
+	}
+}
+
+// The masks are the policy: on a file any write reaches its contents, while on a
+// directory only deleting or taking over an entry reaches something already
+// there. Adding an entry does not, which is why the walk survives a volume root.
+func TestWriteAccessMasks(t *testing.T) {
+	assert.NotZero(t, fileWriteAccess&windows.FILE_WRITE_DATA, "writing a file's data reaches its contents")
+	assert.NotZero(t, fileWriteAccess&windows.FILE_APPEND_DATA, "appending to a file reaches its contents")
+
+	assert.Zero(t, dirWriteAccess&windows.FILE_WRITE_DATA, "adding a file to a directory replaces nothing")
+	assert.Zero(t, dirWriteAccess&windows.FILE_APPEND_DATA, "adding a subdirectory replaces nothing")
+	assert.NotZero(t, dirWriteAccess&fileDeleteChild, "deleting an entry replaces it")
+	assert.NotZero(t, dirWriteAccess&windows.DELETE, "deleting the directory takes its entries with it")
+}
+
+func TestIsAllowACE(t *testing.T) {
+	tests := []struct {
+		name    string
+		aceType uint8
+		want    bool
+	}{
+		{name: "allowed", aceType: windows.ACCESS_ALLOWED_ACE_TYPE, want: true},
+		{name: "allowed callback", aceType: accessAllowedCallbackACEType, want: true},
+		{name: "allowed object", aceType: accessAllowedObjectACEType, want: true},
+		{name: "allowed callback object", aceType: accessAllowedCallbackObjectACEType, want: true},
+		{name: "denied", aceType: windows.ACCESS_DENIED_ACE_TYPE},
+		// SYSTEM_AUDIT_ACE_TYPE, which x/sys does not define: an ACE that records
+		// access rather than granting it.
+		{name: "audit", aceType: 0x2},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			assert.Equal(t, tt.want, isAllowACE(tt.aceType), "ACE type %#x", tt.aceType)
+		})
+	}
+}
+
+// writeExecutable creates a plain file under the test's own directory, the shape
+// trustedSelf checks.
+func writeExecutable(t *testing.T) string {
+	t.Helper()
+	path := filepath.Join(t.TempDir(), "netbird-ui.exe")
+	require.NoError(t, os.WriteFile(path, []byte("MZ"), 0o755), "write the executable")
+	return path
+}
+
+// grantWrite replaces the file's DACL with one that grants a well-known trustee
+// everything, keeping the test user's own access so the file stays deletable.
+func grantWrite(t *testing.T, path string, wellKnown windows.WELL_KNOWN_SID_TYPE) {
+	t.Helper()
+
+	trustee, err := windows.CreateWellKnownSid(wellKnown)
+	require.NoError(t, err, "build the trustee SID")
+	self, err := currentUserSID()
+	require.NoError(t, err, "read the test user's SID")
+
+	acl, err := windows.ACLFromEntries([]windows.EXPLICIT_ACCESS{
+		fullControl(self, windows.TRUSTEE_IS_USER),
+		fullControl(trustee, windows.TRUSTEE_IS_WELL_KNOWN_GROUP),
+	}, nil)
+	require.NoError(t, err, "build the ACL")
+
+	require.NoError(t, windows.SetNamedSecurityInfo(path, windows.SE_FILE_OBJECT,
+		windows.DACL_SECURITY_INFORMATION|windows.PROTECTED_DACL_SECURITY_INFORMATION,
+		nil, nil, acl, nil), "set the DACL")
+}
+
+func fullControl(sid *windows.SID, trusteeType uint32) windows.EXPLICIT_ACCESS {
+	return windows.EXPLICIT_ACCESS{
+		AccessPermissions: windows.GENERIC_ALL,
+		AccessMode:        windows.GRANT_ACCESS,
+		Trustee: windows.TRUSTEE{
+			TrusteeForm:  windows.TRUSTEE_IS_SID,
+			TrusteeType:  windows.TRUSTEE_TYPE(trusteeType),
+			TrusteeValue: windows.TrusteeValueFromSID(sid),
+		},
+	}
+}

client/internal/engine.go

@@ -572,12 +572,7 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 	}
 	e.stateManager.Start()
 
-	initialRoutes, dnsConfig, dnsFeatureFlag, err := e.readInitialSettings()
-	if err != nil {
-		return fmt.Errorf("read initial settings: %w", err)
-	}
-
-	dnsServer, err := e.newDnsServer(dnsConfig)
+	dnsServer, err := e.newDnsServer()
 	if err != nil {
 		return fmt.Errorf("create dns server: %w", err)
 	}
@@ -595,10 +590,8 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 		WGInterface:         e.wgInterface,
 		StatusRecorder:      e.statusRecorder,
 		RelayManager:        e.relayManager,
-		InitialRoutes:       initialRoutes,
 		StateManager:        e.stateManager,
 		DNSServer:           dnsServer,
-		DNSFeatureFlag:      dnsFeatureFlag,
 		PeerStore:           e.peerStore,
 		DisableClientRoutes: e.config.DisableClientRoutes,
 		DisableServerRoutes: e.config.DisableServerRoutes,
@@ -869,7 +862,8 @@ func (e *Engine) modifyPeers(peersUpdate []*mgmProto.RemotePeerConfig) error {
 	}
 	// third, add the peer connections again
 	for _, p := range modified {
-		if err := e.addNewPeer(p); err != nil {
+		err := e.addNewPeer(p)
+		if err != nil {
 			return err
 		}
 	}
@@ -1490,12 +1484,8 @@ func (e *Engine) updateNetworkMap(networkMap *mgmProto.NetworkMap) error {
 		return nil
 	}
 
-	// Only update the flag when the sync carries a peer config; a nil peer config
-	// (e.g. a partial update) must not reset the cached flag to false.
-	if peerConfig := networkMap.GetPeerConfig(); peerConfig != nil {
-		if err := e.connMgr.UpdatedRemoteFeatureFlag(e.ctx, peerConfig.GetLazyConnectionEnabled()); err != nil {
-			log.Errorf("failed to update lazy connection feature flag: %v", err)
-		}
+	if err := e.connMgr.UpdatedRemoteFeatureFlag(e.ctx, networkMap.GetPeerConfig().GetLazyConnectionEnabled()); err != nil {
+		log.Errorf("failed to update lazy connection feature flag: %v", err)
 	}
 
 	if e.firewall != nil {
@@ -1561,7 +1551,8 @@ func (e *Engine) updateNetworkMap(networkMap *mgmProto.NetworkMap) error {
 
 	// Ingress forward rules
 	done = e.phase("forward_rules")
-	if _, err := e.updateForwardRules(networkMap.GetForwardingRules()); err != nil {
+	forwardingRules, err := e.updateForwardRules(networkMap.GetForwardingRules())
+	if err != nil {
 		log.Errorf("failed to update forward rules, err: %v", err)
 	}
 	done()
@@ -1579,7 +1570,8 @@ func (e *Engine) updateNetworkMap(networkMap *mgmProto.NetworkMap) error {
 
 	// must set the exclude list after the peers are added. Without it the manager can not figure out the peers parameters from the store
 	done = e.phase("lazy_exclude")
-	e.connMgr.SetExcludeList(e.ctx, e.toExcludedLazyPeers(remotePeers))
+	excludedLazyPeers := e.toExcludedLazyPeers(forwardingRules, remotePeers)
+	e.connMgr.SetExcludeList(e.ctx, excludedLazyPeers)
 	done()
 
 	e.networkSerial = serial
@@ -1823,15 +1815,15 @@ func addrToString(addr netip.Addr) string {
 // addNewPeers adds peers that were not know before but arrived from the Management service with the update
 func (e *Engine) addNewPeers(peersUpdate []*mgmProto.RemotePeerConfig) error {
 	for _, p := range peersUpdate {
-		if err := e.addNewPeer(p); err != nil {
+		err := e.addNewPeer(p)
+		if err != nil {
 			return err
 		}
 	}
 	return nil
 }
 
-// addNewPeer add peer if connection doesn't exist. A peer that is not lazy by
-// policy gets an always-active connection instead.
+// addNewPeer add peer if connection doesn't exist
 func (e *Engine) addNewPeer(peerConfig *mgmProto.RemotePeerConfig) error {
 	peerKey := peerConfig.GetWgPubKey()
 	peerIPs := make([]netip.Prefix, 0, len(peerConfig.GetAllowedIps()))
@@ -1866,8 +1858,7 @@ func (e *Engine) addNewPeer(peerConfig *mgmProto.RemotePeerConfig) error {
 		log.Warnf("error adding peer %s to status recorder, got error: %v", peerKey, err)
 	}
 
-	permanent := !e.connMgr.PeerLazyDefault(peerConfig.GetLazyState())
-	if exists := e.connMgr.AddPeerConn(e.ctx, peerKey, conn, permanent); exists {
+	if exists := e.connMgr.AddPeerConn(e.ctx, peerKey, conn); exists {
 		conn.Close(false)
 		return fmt.Errorf("peer already exists: %s", peerKey)
 	}
@@ -2104,42 +2095,6 @@ func (e *Engine) close() {
 	}
 }
 
-func (e *Engine) readInitialSettings() ([]*route.Route, *nbdns.Config, bool, error) {
-	if runtime.GOOS != "android" {
-		// nolint:nilnil
-		return nil, nil, false, nil
-	}
-
-	info := system.GetInfo(e.ctx)
-	info.SetFlags(
-		e.config.RosenpassEnabled,
-		e.config.RosenpassPermissive,
-		&e.config.ServerSSHAllowed,
-		e.config.DisableClientRoutes,
-		e.config.DisableServerRoutes,
-		e.config.DisableDNS,
-		e.config.DisableFirewall,
-		e.config.BlockLANAccess,
-		e.config.BlockInbound,
-		e.config.DisableIPv6,
-		e.config.SyncMessageVersion,
-		e.config.EnableSSHRoot,
-		e.config.EnableSSHSFTP,
-		e.config.EnableSSHLocalPortForwarding,
-		e.config.EnableSSHRemotePortForwarding,
-		e.config.DisableSSHAuth,
-	)
-
-	netMap, err := e.mgmClient.GetNetworkMap(info)
-	if err != nil {
-		return nil, nil, false, err
-	}
-	routes := toRoutes(netMap.GetRoutes())
-	dnsCfg := toDNSConfig(netMap.GetDNSConfig(), e.wgInterface.Address())
-	dnsFeatureFlag := toDNSFeatureFlag(netMap)
-	return routes, &dnsCfg, dnsFeatureFlag, nil
-}
-
 func (e *Engine) newWgIface() (*iface.WGIface, error) {
 	transportNet, err := e.newStdNet()
 	if err != nil {
@@ -2174,7 +2129,7 @@ func (e *Engine) newWgIface() (*iface.WGIface, error) {
 func (e *Engine) wgInterfaceCreate() (err error) {
 	switch runtime.GOOS {
 	case "android":
-		err = e.wgInterface.CreateOnAndroid(e.routeManager.InitialRouteRange(), e.dnsServer.DnsIP().String(), e.dnsServer.SearchDomains())
+		err = e.wgInterface.CreateOnAndroid(e.routeManager.CurrentRouteRange(), e.dnsServer.DnsIP().String(), e.dnsServer.SearchDomains())
 	case "ios":
 		e.mobileDep.NetworkChangeListener.SetInterfaceIP(e.config.WgAddr.String())
 		if e.config.WgAddr.HasIPv6() {
@@ -2187,7 +2142,7 @@ func (e *Engine) wgInterfaceCreate() (err error) {
 	return err
 }
 
-func (e *Engine) newDnsServer(dnsConfig *nbdns.Config) (dns.Server, error) {
+func (e *Engine) newDnsServer() (dns.Server, error) {
 	// due to tests where we are using a mocked version of the DNS server
 	if e.dnsServer != nil {
 		return e.dnsServer, nil
@@ -2199,7 +2154,7 @@ func (e *Engine) newDnsServer(dnsConfig *nbdns.Config) (dns.Server, error) {
 			e.ctx,
 			e.wgInterface,
 			e.mobileDep.HostDNSAddresses,
-			*dnsConfig,
+			nbdns.Config{},
 			e.mobileDep.NetworkChangeListener,
 			e.statusRecorder,
 			e.config.DisableDNS,
@@ -2694,19 +2649,46 @@ func (e *Engine) updateForwardRules(rules []*mgmProto.ForwardingRule) ([]firewal
 	return forwardingRules, nberrors.FormatErrorOrNil(merr)
 }
 
-// toExcludedLazyPeers returns the peers that must have an always-active
-// connection: those that are not lazy by policy (the per-peer lazy state or the
-// account flag, subject to the local override).
-func (e *Engine) toExcludedLazyPeers(peers []*mgmProto.RemotePeerConfig) map[string]bool {
+func (e *Engine) toExcludedLazyPeers(rules []firewallManager.ForwardRule, peers []*mgmProto.RemotePeerConfig) map[string]bool {
 	excludedPeers := make(map[string]bool)
-	for _, p := range peers {
-		if !e.connMgr.PeerLazyDefault(p.GetLazyState()) {
-			excludedPeers[p.GetWgPubKey()] = true
+
+	// Ingress forward targets: inbound forwarded traffic is initiated remotely and
+	// cannot wake a lazy connection, so the peer routing the target must stay
+	// permanently connected. AllowedIPs are already parsed on the peer conn, so
+	// reuse those typed prefixes instead of re-parsing the network map strings.
+	for _, r := range rules {
+		for _, p := range peers {
+			if e.peerRoutesAddr(p, r.TranslatedAddress) {
+				log.Infof("exclude forwarder peer from lazy connection: %s", p.GetWgPubKey())
+				excludedPeers[p.GetWgPubKey()] = true
+			}
 		}
 	}
+
 	return excludedPeers
 }
 
+// peerRoutesAddr reports whether the peer is a router for addr, matched against
+// the peer's already-parsed AllowedIPs from the store (the same typed value the
+// lazy manager consumes) rather than re-parsing the network map strings.
+func (e *Engine) peerRoutesAddr(p *mgmProto.RemotePeerConfig, addr netip.Addr) bool {
+	prefixes, ok := e.peerStore.AllowedIPs(p.GetWgPubKey())
+	if !ok {
+		return false
+	}
+	return prefixesContain(prefixes, addr)
+}
+
+// prefixesContain reports whether addr falls within any of the prefixes.
+func prefixesContain(prefixes []netip.Prefix, addr netip.Addr) bool {
+	for _, prefix := range prefixes {
+		if prefix.Contains(addr) {
+			return true
+		}
+	}
+	return false
+}
+
 // isChecksEqual checks if two slices of checks are equal.
 func isChecksEqual(checks1, checks2 []*mgmProto.Checks) bool {
 	normalize := func(checks []*mgmProto.Checks) []string {

client/internal/engine_lazy_exclude_test.go

@@ -0,0 +1,87 @@
+package internal
+
+import (
+	"net/netip"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+
+	firewallManager "github.com/netbirdio/netbird/client/firewall/manager"
+	"github.com/netbirdio/netbird/client/internal/peer"
+	"github.com/netbirdio/netbird/client/internal/peerstore"
+	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
+)
+
+func TestPrefixesContain(t *testing.T) {
+	tests := []struct {
+		name     string
+		prefixes []string
+		addr     string
+		want     bool
+	}{
+		{name: "own overlay /32 matches", prefixes: []string{"100.110.8.145/32"}, addr: "100.110.8.145", want: true},
+		{name: "addr inside routed subnet", prefixes: []string{"10.121.0.0/16"}, addr: "10.121.208.4", want: true},
+		{name: "addr outside subnet", prefixes: []string{"10.121.0.0/16"}, addr: "10.122.0.1", want: false},
+		{name: "different /32", prefixes: []string{"100.110.8.145/32"}, addr: "100.110.8.146", want: false},
+		{name: "ipv6 /128 matches", prefixes: []string{"fd00::1/128"}, addr: "fd00::1", want: true},
+		{name: "no prefixes", prefixes: nil, addr: "10.121.208.4", want: false},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			prefixes := make([]netip.Prefix, 0, len(tt.prefixes))
+			for _, p := range tt.prefixes {
+				prefixes = append(prefixes, netip.MustParsePrefix(p))
+			}
+			require.Equal(t, tt.want, prefixesContain(prefixes, netip.MustParseAddr(tt.addr)))
+		})
+	}
+}
+
+// TestToExcludedLazyPeers_ForwardTarget guards a regression: the forward-target
+// peer (the peer routing a ForwardRule.TranslatedAddress) must be excluded from
+// lazy connections, matched via the peer's already-parsed AllowedIPs.
+func TestToExcludedLazyPeers_ForwardTarget(t *testing.T) {
+	const targetPeerKey = "cccccccccccccccccccccccccccccccccccccccccc0="
+	const otherPeerKey = "dddddddddddddddddddddddddddddddddddddddddd0="
+
+	store := peerstore.NewConnStore()
+	store.AddPeerConn(targetPeerKey, newTestConn(t, targetPeerKey, "100.110.8.145/32"))
+	store.AddPeerConn(otherPeerKey, newTestConn(t, otherPeerKey, "100.110.9.10/32"))
+
+	e := &Engine{peerStore: store}
+
+	peers := []*mgmProto.RemotePeerConfig{
+		{WgPubKey: targetPeerKey, AllowedIps: []string{"100.110.8.145/32"}},
+		{WgPubKey: otherPeerKey, AllowedIps: []string{"100.110.9.10/32"}},
+	}
+	rules := []firewallManager.ForwardRule{
+		{TranslatedAddress: netip.MustParseAddr("100.110.8.145")},
+	}
+
+	excluded := e.toExcludedLazyPeers(rules, peers)
+
+	require.True(t, excluded[targetPeerKey], "forward-target peer must be excluded from lazy connections")
+	require.False(t, excluded[otherPeerKey], "non-target peer must not be excluded")
+	require.Len(t, excluded, 1)
+}
+
+func TestToExcludedLazyPeers_NoRules(t *testing.T) {
+	e := &Engine{peerStore: peerstore.NewConnStore()}
+
+	peers := []*mgmProto.RemotePeerConfig{
+		{WgPubKey: "peer-a", AllowedIps: []string{"100.110.8.145/32"}},
+	}
+
+	require.Empty(t, e.toExcludedLazyPeers(nil, peers))
+}
+
+func newTestConn(t *testing.T, key, allowedIP string) *peer.Conn {
+	t.Helper()
+	conn, err := peer.NewConn(peer.ConnConfig{
+		Key:      key,
+		WgConfig: peer.WgConfig{AllowedIps: []netip.Prefix{netip.MustParsePrefix(allowedIP)}},
+	}, peer.ServiceDependencies{})
+	require.NoError(t, err)
+	return conn
+}

client/internal/engine_test.go

@@ -279,8 +279,7 @@ func TestEngine_UpdateNetworkMap(t *testing.T) {
 	}, MobileDependency{})
 
 	wgIface := &MockWGIface{
-		NameFunc:            func() string { return "utun102" },
-		IsUserspaceBindFunc: func() bool { return true },
+		NameFunc: func() string { return "utun102" },
 		RemovePeerFunc: func(peerKey string) error {
 			return nil
 		},

client/internal/engine_tunsettings.go

@@ -0,0 +1,20 @@
+package internal
+
+func (e *Engine) TunSettings() ([]string, []string) {
+	e.syncMsgMux.Lock()
+	routeManager := e.routeManager
+	dnsServer := e.dnsServer
+	e.syncMsgMux.Unlock()
+
+	var routes []string
+	if routeManager != nil {
+		routes = routeManager.CurrentRouteRange()
+	}
+
+	var searchDomains []string
+	if dnsServer != nil {
+		searchDomains = dnsServer.SearchDomains()
+	}
+
+	return routes, searchDomains
+}

client/internal/ipcauth/privileged.go

@@ -91,6 +91,12 @@ func SelfDelegatesTo() (Identity, bool) {
 	return selfIdentity, true
 }
 
+// The values PrivilegedActorKey returns.
+const (
+	ActorKeyAdministrator = "administrator"
+	ActorKeyRoot          = "root"
+)
+
 // PrivilegedActor names the principal a privileged operation requires, for use
 // in messages shown to the user.
 func PrivilegedActor() string {
@@ -100,6 +106,16 @@ func PrivilegedActor() string {
 	return "root"
 }
 
+// PrivilegedActorKey identifies that principal without wording it, for a client
+// that writes its own message in the user's language. The words PrivilegedActor
+// returns are English, and a translated sentence cannot borrow them.
+func PrivilegedActorKey() string {
+	if runtime.GOOS == "windows" {
+		return ActorKeyAdministrator
+	}
+	return ActorKeyRoot
+}
+
 // ElevatedCommand renders a command so that running it grants the privileges the
 // operation needs. Windows has no in-line equivalent of sudo, so the command is
 // returned unchanged and the user is expected to run it from an elevated

client/internal/profilemanager/state.go

@@ -45,12 +45,35 @@ func (pm *ProfileManager) GetProfileState(id ID) (*ProfileState, error) {
 	return &state, nil
 }
 
-func (pm *ProfileManager) SetActiveProfileState(state *ProfileState) error {
+// SetProfileState writes the state file of the profile identified by id. Prefer
+// it over SetActiveProfileState whenever the caller knows which profile the data
+// belongs to: an SSO login spans seconds of user interaction, and the active
+// profile can change during it, which would file the account email under
+// whichever profile happened to be active when the flow returned.
+func (pm *ProfileManager) SetProfileState(id ID, state *ProfileState) error {
 	configDir, err := getConfigDir()
 	if err != nil {
 		return fmt.Errorf("get config directory: %w", err)
 	}
 
+	if id == "" {
+		return fmt.Errorf("empty profile ID")
+	}
+	if id != defaultProfileName && !IsValidProfileFilenameStem(id) {
+		return fmt.Errorf("invalid profile ID: %q", id)
+	}
+
+	stateFile := filepath.Join(configDir, id.String()+".state.json")
+	if err := util.WriteJsonWithRestrictedPermission(context.Background(), stateFile, state); err != nil {
+		return fmt.Errorf("write profile state: %w", err)
+	}
+
+	return nil
+}
+
+// SetActiveProfileState writes the state file of whichever profile is active at
+// call time. Use SetProfileState when the target profile is known.
+func (pm *ProfileManager) SetActiveProfileState(state *ProfileState) error {
 	activeProf, err := pm.GetActiveProfile()
 	if err != nil {
 		if errors.Is(err, ErrNoActiveProfile) {
@@ -59,18 +82,7 @@ func (pm *ProfileManager) SetActiveProfileState(state *ProfileState) error {
 		return fmt.Errorf("get active profile: %w", err)
 	}
 
-	id := activeProf.ID
-	if id != defaultProfileName && !IsValidProfileFilenameStem(id) {
-		return fmt.Errorf("invalid active profile ID: %q", id)
-	}
-
-	stateFile := filepath.Join(configDir, id.String()+".state.json")
-	err = util.WriteJsonWithRestrictedPermission(context.Background(), stateFile, state)
-	if err != nil {
-		return fmt.Errorf("write profile state: %w", err)
-	}
-
-	return nil
+	return pm.SetProfileState(activeProf.ID, state)
 }
 
 // RemoveProfileState deletes the per-profile state file (which holds the

client/internal/routemanager/dnsinterceptor/handler.go

@@ -479,7 +479,7 @@ func (d *DnsInterceptor) removeDNATMappings(realPrefixes []netip.Prefix, logger
 
 // internalDnatFw checks if the firewall supports internal DNAT
 func (d *DnsInterceptor) internalDnatFw() (internalDNATer, bool) {
-	if d.firewall == nil || runtime.GOOS != "android" {
+	if d.firewall == nil || d.fakeIPManager == nil || runtime.GOOS != "android" {
 		return nil, false
 	}
 	fw, ok := d.firewall.(internalDNATer)

client/internal/routemanager/manager.go

@@ -8,14 +8,13 @@ import (
 	"net/netip"
 	"net/url"
 	"runtime"
-	"slices"
+	"sort"
 	"strings"
 	"sync"
 	"sync/atomic"
 	"syscall"
 	"time"
 
-	"github.com/google/uuid"
 	"github.com/hashicorp/go-multierror"
 	log "github.com/sirupsen/logrus"
 	"golang.org/x/exp/maps"
@@ -62,7 +61,7 @@ type Manager interface {
 	GetActiveClientRoutes() route.HAMap
 	GetClientRoutesWithNetID() map[route.NetID][]*route.Route
 	SetRouteChangeListener(listener listener.NetworkChangeListener)
-	InitialRouteRange() []string
+	CurrentRouteRange() []string
 	SetFirewall(firewall.Manager) error
 	SetDNSForwarderPort(port uint16)
 	ReconcilePeerAllowedIPs(peerKey string) error
@@ -76,10 +75,8 @@ type ManagerConfig struct {
 	WGInterface         iface.WGIface
 	StatusRecorder      *peer.Status
 	RelayManager        *relayClient.Manager
-	InitialRoutes       []*route.Route
 	StateManager        *statemanager.Manager
 	DNSServer           dns.Server
-	DNSFeatureFlag      bool
 	PeerStore           *peerstore.Store
 	DisableClientRoutes bool
 	DisableServerRoutes bool
@@ -149,45 +146,12 @@ func NewManager(config ManagerConfig) *DefaultManager {
 	useNoop := netstack.IsEnabled() || config.DisableClientRoutes
 	dm.setupRefCounters(useNoop)
 
-	// don't proceed with client routes if it is disabled
-	if config.DisableClientRoutes {
-		return dm
-	}
-
-	if runtime.GOOS == "android" {
-		dm.setupAndroidRoutes(config)
-	}
 	return dm
 }
-func (m *DefaultManager) setupAndroidRoutes(config ManagerConfig) {
-	cr := m.initialClientRoutes(config.InitialRoutes)
 
-	routesForComparison := slices.Clone(cr)
-
-	if config.DNSFeatureFlag {
-		m.fakeIPManager = fakeip.NewManager()
-
-		v4ID := uuid.NewString()
-		fakeIPRoute := &route.Route{
-			ID:          route.ID(v4ID),
-			Network:     m.fakeIPManager.GetFakeIPBlock(),
-			NetID:       route.NetID(v4ID),
-			Peer:        m.pubKey,
-			NetworkType: route.IPv4Network,
-		}
-		v6ID := uuid.NewString()
-		fakeIPv6Route := &route.Route{
-			ID:          route.ID(v6ID),
-			Network:     m.fakeIPManager.GetFakeIPv6Block(),
-			NetID:       route.NetID(v6ID),
-			Peer:        m.pubKey,
-			NetworkType: route.IPv6Network,
-		}
-		cr = append(cr, fakeIPRoute, fakeIPv6Route)
-		m.notifier.SetFakeIPRoutes([]*route.Route{fakeIPRoute, fakeIPv6Route})
-	}
-
-	m.notifier.SetInitialClientRoutes(cr, routesForComparison)
+func (m *DefaultManager) enableFakeIPRoutes() {
+	m.fakeIPManager = fakeip.NewManager()
+	m.notifier.NotifyRouteChange()
 }
 
 func (m *DefaultManager) setupRefCounters(useNoop bool) {
@@ -464,6 +428,9 @@ func (m *DefaultManager) UpdateRoutes(
 
 	var merr *multierror.Error
 	if !m.disableClientRoutes {
+		if runtime.GOOS == "android" && useNewDNSRoute && m.fakeIPManager == nil {
+			m.enableFakeIPRoutes()
+		}
 
 		// Update route selector based on management server's isSelected status
 		m.updateRouteSelectorFromManagement(clientRoutes)
@@ -500,9 +467,32 @@ func (m *DefaultManager) SetRouteChangeListener(listener listener.NetworkChangeL
 	m.notifier.SetListener(listener)
 }
 
-// InitialRouteRange return the list of initial routes. It used by mobile systems
-func (m *DefaultManager) InitialRouteRange() []string {
-	return m.notifier.GetInitialRouteRanges()
+// CurrentRouteRange returns the current TUN route list. It is used by mobile systems
+func (m *DefaultManager) CurrentRouteRange() []string {
+	m.mux.Lock()
+	defer m.mux.Unlock()
+
+	if m.disableClientRoutes {
+		return nil
+	}
+
+	filtered := m.routeSelector.FilterSelectedExitNodes(m.clientRoutes)
+	var nets []string
+	for _, routes := range filtered {
+		for _, r := range routes {
+			if r.IsDynamic() {
+				continue
+			}
+			nets = append(nets, r.NetString())
+		}
+	}
+
+	if m.fakeIPManager != nil {
+		nets = append(nets, m.fakeIPManager.GetFakeIPBlock().String(), m.fakeIPManager.GetFakeIPv6Block().String())
+	}
+
+	sort.Strings(nets)
+	return nets
 }
 
 // GetRouteSelector returns the route selector
@@ -700,16 +690,6 @@ func (m *DefaultManager) ClassifyRoutes(newRoutes []*route.Route) (map[route.ID]
 	return newServerRoutesMap, newClientRoutesIDMap
 }
 
-func (m *DefaultManager) initialClientRoutes(initialRoutes []*route.Route) []*route.Route {
-	_, crMap := m.ClassifyRoutes(initialRoutes)
-	rs := make([]*route.Route, 0, len(crMap))
-	for _, routes := range crMap {
-		rs = append(rs, routes...)
-	}
-
-	return rs
-}
-
 func isRouteSupported(route *route.Route) bool {
 	if netstack.IsEnabled() || !nbnet.CustomRoutingDisabled() || route.IsDynamic() {
 		return true

client/internal/routemanager/mock.go

@@ -30,8 +30,8 @@ func (m *MockManager) Init() error {
 	return nil
 }
 
-// InitialRouteRange mock implementation of InitialRouteRange from Manager interface
-func (m *MockManager) InitialRouteRange() []string {
+// CurrentRouteRange mock implementation of CurrentRouteRange from Manager interface
+func (m *MockManager) CurrentRouteRange() []string {
 	return nil
 }
 

client/internal/routemanager/notifier/notifier_android.go

@@ -6,7 +6,6 @@ import (
 	"net/netip"
 	"slices"
 	"sort"
-	"strings"
 	"sync"
 
 	"github.com/netbirdio/netbird/client/internal/listener"
@@ -14,12 +13,15 @@ import (
 )
 
 type Notifier struct {
-	initialRoutes []*route.Route
-	currentRoutes []*route.Route
-	fakeIPRoutes  []*route.Route
+	mu sync.Mutex
 
-	listener    listener.NetworkChangeListener
-	listenerMux sync.Mutex
+	// currentRoutes is the last announced route set. It exists only to
+	// suppress noise: without it every network map sync would trigger the
+	// Java side, even when the routes did not change. The actual TUN route
+	// state is owned by the route manager and pulled from there.
+	currentRoutes []*route.Route
+
+	listener listener.NetworkChangeListener
 }
 
 func NewNotifier() *Notifier {
@@ -27,20 +29,15 @@ func NewNotifier() *Notifier {
 }
 
 func (n *Notifier) SetListener(listener listener.NetworkChangeListener) {
-	n.listenerMux.Lock()
-	defer n.listenerMux.Unlock()
+	n.mu.Lock()
+	defer n.mu.Unlock()
 	n.listener = listener
 }
 
-// SetInitialClientRoutes stores the initial route sets for TUN configuration.
-func (n *Notifier) SetInitialClientRoutes(initialRoutes []*route.Route, routesForComparison []*route.Route) {
-	n.initialRoutes = filterStatic(initialRoutes)
-	n.currentRoutes = filterStatic(routesForComparison)
-}
-
-// SetFakeIPRoutes stores the fake IP routes to be included in every TUN rebuild.
-func (n *Notifier) SetFakeIPRoutes(routes []*route.Route) {
-	n.fakeIPRoutes = routes
+func (n *Notifier) NotifyRouteChange() {
+	n.mu.Lock()
+	defer n.mu.Unlock()
+	n.notifyLocked()
 }
 
 func (n *Notifier) OnNewRoutes(idMap route.HAMap) {
@@ -54,46 +51,32 @@ func (n *Notifier) OnNewRoutes(idMap route.HAMap) {
 		}
 	}
 
-	if !n.hasRouteDiff(n.currentRoutes, newRoutes) {
+	n.mu.Lock()
+	defer n.mu.Unlock()
+	if !hasRouteDiff(n.currentRoutes, newRoutes) {
 		return
 	}
 
 	n.currentRoutes = newRoutes
-	n.notify()
+	n.notifyLocked()
 }
 
 func (n *Notifier) OnNewPrefixes([]netip.Prefix) {
 	// Not used on Android
 }
 
-func (n *Notifier) notify() {
-	n.listenerMux.Lock()
-	defer n.listenerMux.Unlock()
+func (n *Notifier) notifyLocked() {
 	if n.listener == nil {
 		return
 	}
-
-	allRoutes := slices.Clone(n.currentRoutes)
-	allRoutes = append(allRoutes, n.fakeIPRoutes...)
-
-	routeStrings := n.routesToStrings(allRoutes)
-	sort.Strings(routeStrings)
-	go func(l listener.NetworkChangeListener) {
-		l.OnNetworkChanged(strings.Join(routeStrings, ","))
-	}(n.listener)
+	n.listener.OnNetworkChanged("")
 }
 
-func filterStatic(routes []*route.Route) []*route.Route {
-	out := make([]*route.Route, 0, len(routes))
-	for _, r := range routes {
-		if !r.IsDynamic() {
-			out = append(out, r)
-		}
-	}
-	return out
+func (n *Notifier) Close() {
+	// unused
 }
 
-func (n *Notifier) routesToStrings(routes []*route.Route) []string {
+func routesToStrings(routes []*route.Route) []string {
 	nets := make([]string, 0, len(routes))
 	for _, r := range routes {
 		nets = append(nets, r.NetString())
@@ -101,25 +84,10 @@ func (n *Notifier) routesToStrings(routes []*route.Route) []string {
 	return nets
 }
 
-func (n *Notifier) hasRouteDiff(a []*route.Route, b []*route.Route) bool {
-	slices.SortFunc(a, func(x, y *route.Route) int {
-		return strings.Compare(x.NetString(), y.NetString())
-	})
-	slices.SortFunc(b, func(x, y *route.Route) int {
-		return strings.Compare(x.NetString(), y.NetString())
-	})
-
-	return !slices.EqualFunc(a, b, func(x, y *route.Route) bool {
-		return x.NetString() == y.NetString()
-	})
-}
-
-func (n *Notifier) GetInitialRouteRanges() []string {
-	initialStrings := n.routesToStrings(n.initialRoutes)
-	sort.Strings(initialStrings)
-	return initialStrings
-}
-
-func (n *Notifier) Close() {
-	// unused
+func hasRouteDiff(a []*route.Route, b []*route.Route) bool {
+	as := routesToStrings(a)
+	bs := routesToStrings(b)
+	sort.Strings(as)
+	sort.Strings(bs)
+	return !slices.Equal(as, bs)
 }

client/internal/routemanager/notifier/notifier_ios.go

@@ -29,11 +29,7 @@ func (n *Notifier) SetListener(listener listener.NetworkChangeListener) {
 	n.listener = listener
 }
 
-func (n *Notifier) SetInitialClientRoutes([]*route.Route, []*route.Route) {
-	// iOS doesn't care about initial routes
-}
-
-func (n *Notifier) SetFakeIPRoutes([]*route.Route) {
+func (n *Notifier) NotifyRouteChange() {
 	// Not used on iOS
 }
 

client/internal/routemanager/notifier/notifier_other.go

@@ -19,11 +19,7 @@ func (n *Notifier) SetListener(listener listener.NetworkChangeListener) {
 	// Not used on non-mobile platforms
 }
 
-func (n *Notifier) SetInitialClientRoutes([]*route.Route, []*route.Route) {
-	// Not used on non-mobile platforms
-}
-
-func (n *Notifier) SetFakeIPRoutes([]*route.Route) {
+func (n *Notifier) NotifyRouteChange() {
 	// Not used on non-mobile platforms
 }
 
@@ -35,10 +31,6 @@ func (n *Notifier) OnNewPrefixes(prefixes []netip.Prefix) {
 	// Not used on non-mobile platforms
 }
 
-func (n *Notifier) GetInitialRouteRanges() []string {
-	return []string{}
-}
-
 func (n *Notifier) Close() {
 	// unused
 }

client/internal/updater/installer/installer_run_darwin.go

@@ -98,47 +98,44 @@ func (u *Installer) startDaemon(daemonFolder string) error {
 func (u *Installer) startUIAsUser() error {
 	log.Infof("starting netbird-ui: %s", uiBinary)
 
-	// Get the current console user
-	cmd := exec.Command("stat", "-f", "%Su", "/dev/console")
-	output, err := cmd.Output()
+	username, err := consoleUser()
 	if err != nil {
-		return fmt.Errorf("failed to get console user: %w", err)
+		return err
 	}
 
-	username := strings.TrimSpace(string(output))
-	if username == "" || username == "root" {
-		return fmt.Errorf("no active user session found")
-	}
-
-	log.Infof("starting UI for user: %s", username)
-
-	// Get user's UID
 	userInfo, err := user.Lookup(username)
 	if err != nil {
-		return fmt.Errorf("failed to lookup user %s: %w", username, err)
+		return fmt.Errorf("lookup user %s: %w", username, err)
 	}
 
-	// Start the UI process as the console user using launchctl
-	// This ensures the app runs in the user's context with proper GUI access
-	launchCmd := exec.Command("launchctl", "asuser", userInfo.Uid, "open", "-a", uiBinary)
+	log.Infof("starting UI for user: %s (uid %s)", username, userInfo.Uid)
+
+	launchCmd := exec.Command("launchctl", "asuser", userInfo.Uid, "sudo", "-u", username, "-H", "open", "-a", uiBinary)
 	log.Infof("launchCmd: %s", launchCmd.String())
-	// Set the user's home directory for proper macOS app behavior
-	launchCmd.Env = append(os.Environ(), "HOME="+userInfo.HomeDir)
-	log.Infof("set HOME environment variable: %s", userInfo.HomeDir)
 
-	if err := launchCmd.Start(); err != nil {
-		return fmt.Errorf("failed to start UI process: %w", err)
-	}
-
-	// Release the process so it can run independently
-	if err := launchCmd.Process.Release(); err != nil {
-		log.Warnf("failed to release UI process: %v", err)
+	if err := launchCmd.Run(); err != nil {
+		return fmt.Errorf("run UI launch: %w", err)
 	}
 
 	log.Infof("netbird-ui started successfully for user %s", username)
 	return nil
 }
 
+func consoleUser() (string, error) {
+	output, err := exec.Command("stat", "-f", "%Su", "/dev/console").Output()
+	if err != nil {
+		return "", fmt.Errorf("get console user: %w", err)
+	}
+
+	username := strings.TrimSpace(string(output))
+	switch username {
+	case "", "root", "loginwindow", "_mbsetupuser":
+		return "", fmt.Errorf("no active GUI user session, console user: %q", username)
+	}
+
+	return username, nil
+}
+
 func (u *Installer) installPkgFile(ctx context.Context, path string) error {
 	log.Infof("installing pkg file: %s", path)
 

client/ios/NetBirdSDK/client.go

@@ -158,13 +158,19 @@ func (c *Client) Run(fd int32, interfaceName string, envList *EnvList) error {
 	defer c.ctxCancel()
 	c.ctxCancelLock.Unlock()
 
-	auth := NewAuthWithConfig(ctx, cfg)
-	err = auth.LoginSync()
-	if err != nil {
-		return err
-	}
-
-	log.Infof("Auth successful")
+	// No login pre-flight here. The engine's own loginToManagement (connect.go) performs
+	// the authoritative Login immediately before the first Sync, so a LoginSync() call at
+	// this point only duplicated it — costing two extra Login RPCs (IsLoginRequired +
+	// Login) on every engine start, since IsLoginRequired is itself a full Login RPC.
+	//
+	// Auth failures still reach the caller through the engine path: loginToManagement
+	// returns PermissionDenied, which marks the shared status recorder
+	// (MarkManagementDisconnected) and fires ClientStop → onDisconnected, where
+	// IsLoginRequiredCached() reports login-required. The error is also returned out of Run().
+	//
+	// A pre-flight was also actively harmful when the server is unreachable: its 2-minute
+	// backoff blocked the start and then reported "login required" for what was really a
+	// timeout. The engine instead keeps retrying and recovers when the server returns.
 	// todo do not throw error in case of cancelled context
 	ctx = internal.CtxInitState(ctx)
 	c.onHostDnsFn = func([]string) {}

client/ios/NetBirdSDK/login.go

@@ -222,17 +222,36 @@ func (a *Auth) Login(resultListener ErrListener, urlOpener URLOpener, forceDevic
 // LoginWithDeviceName performs interactive login with device authentication support
 // The deviceName parameter allows specifying a custom device name (required for tvOS)
 func (a *Auth) LoginWithDeviceName(resultListener ErrListener, urlOpener URLOpener, forceDeviceAuth bool, deviceName string) {
+	a.startLogin(resultListener, urlOpener, forceDeviceAuth, deviceName, false)
+}
+
+// LoginInteractive performs the same interactive login as LoginWithDeviceName but skips the
+// IsLoginRequired() pre-flight and goes straight to the browser / device-code flow.
+//
+// IsLoginRequired() is itself a full Login RPC against the management server, so when the
+// caller has ALREADY established that login is required it is a pure duplicate. On iOS the
+// main app decides to show the browser based on its own isLoginRequired() check and then
+// calls straight into this method, so re-asking the server would add another Login RPC to
+// every interactive login.
+//
+// Use LoginWithDeviceName when the auth state is unknown and a silent (browser-less) login
+// must still be possible; use this when the browser is going to be shown regardless.
+func (a *Auth) LoginInteractive(resultListener ErrListener, urlOpener URLOpener, forceDeviceAuth bool, deviceName string) {
+	a.startLogin(resultListener, urlOpener, forceDeviceAuth, deviceName, true)
+}
+
+func (a *Auth) startLogin(resultListener ErrListener, urlOpener URLOpener, forceDeviceAuth bool, deviceName string, skipLoginCheck bool) {
 	if resultListener == nil {
-		log.Errorf("LoginWithDeviceName: resultListener is nil")
+		log.Errorf("startLogin: resultListener is nil")
 		return
 	}
 	if urlOpener == nil {
-		log.Errorf("LoginWithDeviceName: urlOpener is nil")
+		log.Errorf("startLogin: urlOpener is nil")
 		resultListener.OnError(fmt.Errorf("urlOpener is nil"))
 		return
 	}
 	go func() {
-		err := a.login(urlOpener, forceDeviceAuth, deviceName)
+		err := a.login(urlOpener, forceDeviceAuth, deviceName, skipLoginCheck)
 		if err != nil {
 			resultListener.OnError(err)
 		} else {
@@ -241,7 +260,7 @@ func (a *Auth) LoginWithDeviceName(resultListener ErrListener, urlOpener URLOpen
 	}()
 }
 
-func (a *Auth) login(urlOpener URLOpener, forceDeviceAuth bool, deviceName string) error {
+func (a *Auth) login(urlOpener URLOpener, forceDeviceAuth bool, deviceName string, skipLoginCheck bool) error {
 	// Create context with device name if provided
 	ctx := a.ctx
 	if deviceName != "" {
@@ -255,10 +274,13 @@ func (a *Auth) login(urlOpener URLOpener, forceDeviceAuth bool, deviceName strin
 	}
 	defer authClient.Close()
 
-	// check if we need to generate JWT token
-	needsLogin, err := authClient.IsLoginRequired(ctx)
-	if err != nil {
-		return fmt.Errorf("failed to check login requirement: %v", err)
+	// check if we need to generate JWT token (skipped when the caller already knows)
+	needsLogin := true
+	if !skipLoginCheck {
+		needsLogin, err = authClient.IsLoginRequired(ctx)
+		if err != nil {
+			return fmt.Errorf("failed to check login requirement: %v", err)
+		}
 	}
 
 	jwtToken := ""

client/server/login_outcome_test.go

@@ -0,0 +1,110 @@
+package server
+
+import (
+	"context"
+	"encoding/json"
+	"errors"
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/internal"
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+// A login that never reached Management is not a decision about the peer's
+// credentials, so it must come back as a retryable error rather than an SSO
+// prompt: the user cannot finish a browser login while Management is down, and
+// the CLI's own backoff resolves the outage on its own once the daemon reports
+// the failure. Reproduces `netbird down; netbird up` printing a device-code URL
+// because Management happened to be restarting when the daemon dialed it.
+func TestLogin_ManagementUnreachableIsReturnedInsteadOfDemandingSSO(t *testing.T) {
+	s, _, _, username, _ := setupServerWithProfile(t)
+	s.rootCtx = internal.CtxInitState(context.Background())
+
+	unreachable := errors.New("create connection: dial context: context deadline exceeded")
+	attempts := 0
+	s.isLoginRequiredFn = func(context.Context) (bool, error) {
+		attempts++
+		return false, unreachable
+	}
+
+	resp, err := s.Login(userCtx(), &proto.LoginRequest{Username: &username})
+	require.Error(t, err)
+	require.ErrorIs(t, err, unreachable, "the transport failure was replaced by something else")
+	require.Nil(t, resp, "a failed login must not answer with a login response")
+	require.Equal(t, 1, attempts)
+	require.Nil(t, s.oauthAuthFlow.flow, "the daemon started an SSO flow for a peer whose login was never decided")
+
+	status, err := internal.CtxGetState(s.rootCtx).Status()
+	require.NoError(t, err)
+	require.Equal(t, internal.StatusLoginFailed, status,
+		"a peer that could not reach Management is not waiting on a login")
+}
+
+// The counterpart: Management refusing the peer's credentials is a decision, and
+// the SSO flow still has to start for it. The profile carries an unusable
+// private key so the flow setup fails immediately instead of dialing, which is
+// enough to show the branch was entered — the refusal itself is never what comes
+// back out.
+func TestLogin_AuthRefusalStartsSSOFlow(t *testing.T) {
+	s, _, _, username, cfgPath := setupServerWithProfile(t)
+	s.rootCtx = internal.CtxInitState(context.Background())
+	breakProfilePrivateKey(t, cfgPath)
+
+	s.isLoginRequiredFn = func(context.Context) (bool, error) {
+		return true, nil
+	}
+
+	_, err := s.Login(userCtx(), &proto.LoginRequest{Username: &username})
+	require.Error(t, err)
+
+	status, stateErr := internal.CtxGetState(s.rootCtx).Status()
+	require.NoError(t, stateErr)
+	require.Equal(t, internal.StatusLoginFailed, status,
+		"the SSO flow setup was never reached with the broken key")
+}
+
+func TestLogin_SetupKeyStillRunsWhenPeerNeedsLogin(t *testing.T) {
+	s, _, _, username, _ := setupServerWithProfile(t)
+	s.rootCtx = internal.CtxInitState(context.Background())
+
+	s.isLoginRequiredFn = func(context.Context) (bool, error) {
+		return true, nil
+	}
+
+	var keysTried []string
+	s.loginAttemptFn = func(_ context.Context, setupKey, _ string) (internal.StatusType, error) {
+		keysTried = append(keysTried, setupKey)
+		return "", nil
+	}
+
+	setupKey := "A2C8E32F-AEB2-4B45-8FD3-8A0C1B2D3E4F"
+	resp, err := s.Login(userCtx(), &proto.LoginRequest{Username: &username, SetupKey: setupKey})
+	require.NoError(t, err, "the probe's outcome leaked out as the login result")
+	require.NotNil(t, resp)
+	require.Equal(t, []string{setupKey}, keysTried, "the setup key never reached the login attempt")
+	require.Nil(t, s.oauthAuthFlow.flow, "a setup-key login started an SSO flow")
+
+	status, err := internal.CtxGetState(s.rootCtx).Status()
+	require.NoError(t, err)
+	require.Equal(t, internal.StatusIdle, status)
+}
+
+// breakProfilePrivateKey replaces the profile's private key with an unparseable
+// one, which makes any attempt to build a Management client fail on the spot.
+func breakProfilePrivateKey(t *testing.T, cfgPath string) {
+	t.Helper()
+
+	raw, err := os.ReadFile(cfgPath)
+	require.NoError(t, err)
+
+	var cfg map[string]any
+	require.NoError(t, json.Unmarshal(raw, &cfg))
+	cfg["PrivateKey"] = "not-a-key"
+
+	patched, err := json.Marshal(cfg)
+	require.NoError(t, err)
+	require.NoError(t, os.WriteFile(cfgPath, patched, 0o600))
+}

client/server/server.go

@@ -135,6 +135,13 @@ type Server struct {
 	updateManager *updater.Manager
 
 	jwtCache *jwtCache
+
+	// loginAttemptFn stands in for the Management login round trip. Tests set
+	// it to drive the login outcomes that need a server on the other end;
+	// production leaves it nil, and every login goes through loginAttempt.
+	loginAttemptFn func(ctx context.Context, setupKey, jwtToken string) (internal.StatusType, error)
+
+	isLoginRequiredFn func(ctx context.Context) (bool, error)
 }
 
 type oauthAuthFlow struct {
@@ -370,7 +377,34 @@ func (s *Server) connectionGoroutineRunning() bool {
 	}
 }
 
-// loginAttempt attempts to login using the provided information. it returns a status in case something fails
+// attemptLogin runs a login round trip against Management, or the stand-in a
+// test installed in place of it.
+func (s *Server) attemptLogin(ctx context.Context, setupKey, jwtToken string) (internal.StatusType, error) {
+	if s.loginAttemptFn != nil {
+		return s.loginAttemptFn(ctx, setupKey, jwtToken)
+	}
+	return s.loginAttempt(ctx, setupKey, jwtToken)
+}
+
+func (s *Server) isLoginRequired(ctx context.Context) (bool, error) {
+	if s.isLoginRequiredFn != nil {
+		return s.isLoginRequiredFn(ctx)
+	}
+
+	authClient, err := auth.NewAuth(ctx, s.config.PrivateKey, s.config.ManagementURL, s.config)
+	if err != nil {
+		log.Errorf("failed to create auth client: %v", err)
+		return false, err
+	}
+	defer authClient.Close()
+
+	return authClient.IsLoginRequired(ctx)
+}
+
+// loginAttempt attempts to login using the provided information. It returns
+// StatusNeedsLogin when Management refused the peer's credentials and
+// StatusLoginFailed for every other failure, so callers can tell an
+// authentication decision apart from a login that never got made.
 func (s *Server) loginAttempt(ctx context.Context, setupKey, jwtToken string) (internal.StatusType, error) {
 	authClient, err := auth.NewAuth(ctx, s.config.PrivateKey, s.config.ManagementURL, s.config)
 	if err != nil {
@@ -623,7 +657,19 @@ func (s *Server) Login(callerCtx context.Context, msg *proto.LoginRequest) (*pro
 	s.config = config
 	s.mutex.Unlock()
 
-	if _, err := s.loginAttempt(ctx, "", ""); err == nil {
+	// A probe that errors leaves the login undecided: Management unreachable, a
+	// restart mid-request, an internal error. Those are returned for the caller
+	// to retry, because turning them into an SSO prompt asks the user to solve
+	// something that is not theirs to solve, and a browser login cannot succeed
+	// while Management is unreachable anyway. Only Management refusing the
+	// peer's key is a decision, and IsLoginRequired reports that as
+	// needsLogin=true rather than an error.
+	needsLogin, err := s.isLoginRequired(ctx)
+	if err != nil {
+		state.Set(internal.StatusLoginFailed)
+		return nil, err
+	}
+	if !needsLogin {
 		state.Set(internal.StatusIdle)
 		return &proto.LoginResponse{}, nil
 	}
@@ -684,7 +730,7 @@ func (s *Server) Login(callerCtx context.Context, msg *proto.LoginRequest) (*pro
 	// which returns NeedsLogin and parks on the browser leg.
 	state.Set(internal.StatusConnecting)
 
-	if loginStatus, err := s.loginAttempt(ctx, msg.SetupKey, ""); err != nil {
+	if loginStatus, err := s.attemptLogin(ctx, msg.SetupKey, ""); err != nil {
 		state.Set(loginStatus)
 		return nil, err
 	}
@@ -839,7 +885,7 @@ func (s *Server) WaitSSOLogin(callerCtx context.Context, msg *proto.WaitSSOLogin
 	s.oauthAuthFlow.expiresAt = time.Now()
 	s.mutex.Unlock()
 
-	if loginStatus, err := s.loginAttempt(ctx, "", tokenInfo.GetTokenToUse()); err != nil {
+	if loginStatus, err := s.attemptLogin(ctx, "", tokenInfo.GetTokenToUse()); err != nil {
 		state.Set(loginStatus)
 		return nil, err
 	}
@@ -1769,6 +1815,9 @@ func (s *Server) RequestExtendAuthSession(
 	if connectClient == nil {
 		return nil, gstatus.Errorf(codes.FailedPrecondition, "client is not running")
 	}
+	if connectClient.Engine() == nil {
+		return nil, gstatus.Errorf(codes.FailedPrecondition, "session can no longer be extended, log in again to reconnect")
+	}
 
 	hint := ""
 	if msg.Hint != nil {

client/ui/build/linux/netbird-ui.desktop

@@ -1,10 +1,11 @@
 [Desktop Entry]
 Type=Application
-Name=netbird-ui
+Name=NetBird
+Comment=NetBird desktop client
 Exec=env WEBKIT_DISABLE_DMABUF_RENDERER=1 netbird-ui
 Icon=netbird-ui
-Categories=Development;
+Categories=Utility;Network;
 Terminal=false
-Keywords=wails
+Keywords=netbird;vpn;wireguard;
 Version=1.0
 StartupNotify=false

client/ui/build/linux/netbird.desktop

@@ -1,5 +1,6 @@
 [Desktop Entry]
-Name=Netbird
+Name=NetBird
+Comment=NetBird desktop client
 Exec=env WEBKIT_DISABLE_DMABUF_RENDERER=1 /usr/bin/netbird-ui
 Icon=netbird
 Type=Application

client/ui/build/linux/nfpm/nfpm.yaml

@@ -21,22 +21,31 @@ contents:
     dst: "/usr/local/bin/netbird-ui"
   - src: "./build/appicon.png"
     dst: "/usr/share/icons/hicolor/128x128/apps/netbird-ui.png"
+  # The name the polkit action's icon_name refers to, which the released packages
+  # install as /usr/share/pixmaps/netbird.png.
+  - src: "./build/appicon.png"
+    dst: "/usr/share/icons/hicolor/128x128/apps/netbird.png"
   - src: "./build/linux/netbird-ui.desktop"
     dst: "/usr/share/applications/netbird-ui.desktop"
+  # Names the polkit action for the elevation prompt the app raises when an
+  # unprivileged user changes a privileged setting; without it the dialog shows a
+  # raw command line.
+  - src: "./build/linux/polkit/io.netbird.settings.policy"
+    dst: "/usr/share/polkit-1/actions/io.netbird.settings.policy"
 
 # Default dependencies for the GTK4 + WebKitGTK 6.0 stack (Ubuntu 24.04+ / Debian 13+)
 depends:
-  - libgtk-4-1
+  - libgtk-4-1 (>= 4.14)
   - libwebkitgtk-6.0-4
   - xdg-utils
 
 # Distribution-specific overrides for different package formats
 overrides:
-  # RPM packages for Fedora / RHEL / AlmaLinux / Rocky Linux
+  # RPM packages for Fedora / RHEL / AlmaLinux / Rocky Linux / openSUSE
   rpm:
     depends:
-      - gtk4
-      - webkitgtk6.0
+      - (gtk4 >= 4.14 or libgtk-4-1 >= 4.14)
+      - (webkitgtk6.0 or libwebkitgtk-6_0-4)
       - xdg-utils
 
   # Arch Linux packages

client/ui/build/linux/polkit/io.netbird.settings.policy

@@ -0,0 +1,47 @@
+<?xml version="1.0" encoding="UTF-8"?>
+<!DOCTYPE policyconfig PUBLIC "-//freedesktop//DTD PolicyKit Policy Configuration 1.0//EN"
+ "http://www.freedesktop.org/standards/PolicyKit/1/policyconfig.dtd">
+
+<!--
+  Names the action behind the elevation prompt the desktop app raises for an SSH
+  setting the daemon restricts to root; without it pkexec's generic dialog offers
+  the raw command line instead. The argv1 annotation keeps this wording to the
+  one-shot mode that applies those settings.
+
+  auth_admin rather than auth_admin_keep: each of these settings is its own grant
+  of shell access, so a credential cache would let a second, unasked-for change
+  ride along on the authorization given the first.
+
+  exec.path takes no wildcard and the binary's location depends on the package,
+  hence one action per path.
+-->
+<policyconfig>
+  <vendor>NetBird</vendor>
+  <vendor_url>https://netbird.io</vendor_url>
+
+  <action id="io.netbird.settings.apply-privileged">
+    <description>Change privileged NetBird settings</description>
+    <message>Authentication is required to change NetBird settings that grant SSH access to this computer.</message>
+    <icon_name>netbird</icon_name>
+    <defaults>
+      <allow_any>auth_admin</allow_any>
+      <allow_inactive>auth_admin</allow_inactive>
+      <allow_active>auth_admin</allow_active>
+    </defaults>
+    <annotate key="org.freedesktop.policykit.exec.path">/usr/bin/netbird-ui</annotate>
+    <annotate key="org.freedesktop.policykit.exec.argv1">--apply-privileged-settings</annotate>
+  </action>
+
+  <action id="io.netbird.settings.apply-privileged-local">
+    <description>Change privileged NetBird settings</description>
+    <message>Authentication is required to change NetBird settings that grant SSH access to this computer.</message>
+    <icon_name>netbird</icon_name>
+    <defaults>
+      <allow_any>auth_admin</allow_any>
+      <allow_inactive>auth_admin</allow_inactive>
+      <allow_active>auth_admin</allow_active>
+    </defaults>
+    <annotate key="org.freedesktop.policykit.exec.path">/usr/local/bin/netbird-ui</annotate>
+    <annotate key="org.freedesktop.policykit.exec.argv1">--apply-privileged-settings</annotate>
+  </action>
+</policyconfig>

client/ui/frontend/src/contexts/SettingsContext.tsx

@@ -22,12 +22,18 @@ const logSaveError = (err: unknown) => console.error("[SettingsContext] save fai
 
 export type AutostartState = { supported: boolean; enabled: boolean };
 
+// GuardedField is a setting the daemon only accepts from root/administrator.
+// Turning one on goes through saveGuardedField, which asks the operating system
+// for the privileges rather than sending a request that would be refused.
+export type GuardedField = "serverSshAllowed" | "enableSshRoot" | "disableSshAuth";
+
 type SettingsContextValue = {
     config: Config;
     guiVersion: string;
     setField: <K extends keyof Config>(k: K, v: Config[K]) => void;
     saveField: <K extends keyof Config>(k: K, v: Config[K]) => Promise<void>;
     saveFields: (partial: Partial<Config>, opts?: { preSharedKey?: string }) => Promise<void>;
+    saveGuardedField: (k: GuardedField, v: boolean) => Promise<void>;
     saveNow: () => Promise<void>;
 };
 
@@ -63,6 +69,12 @@ const useSettingsState = () => {
     const [guiVersion, setGuiVersion] = useState<string>("—");
     const saveTimer = useRef<ReturnType<typeof setTimeout> | null>(null);
     const loadedRef = useRef<LoadedConfig | null>(null);
+    // Set when the daemon's config changed while a save was pending, so the read
+    // that was skipped to protect the pending edit happens once it is through.
+    // Without it the form keeps values the daemon no longer has and the next save
+    // submits them, which for a guarded setting means asking the user to authorize
+    // a change they never made.
+    const reloadOwed = useRef(false);
 
     useEffect(() => {
         loadedRef.current = loaded;
@@ -73,6 +85,7 @@ const useSettingsState = () => {
     // update the daemon then rejected.
     const reload = useCallback(
         async (profileName: string) => {
+            reloadOwed.current = false;
             try {
                 const data = await SettingsSvc.GetConfig({ profileName, username });
                 setLoaded({ profileName, data });
@@ -94,7 +107,12 @@ const useSettingsState = () => {
                     username,
                 });
                 if (cancelled) return;
-                if (saveTimer.current) return;
+                // A pending edit outranks the daemon's copy until it is saved, so
+                // the read is owed rather than dropped: see reloadOwed.
+                if (saveTimer.current) {
+                    reloadOwed.current = true;
+                    return;
+                }
                 setLoaded({ profileName: activeProfileId, data });
             } catch (e) {
                 if (cancelled || !showError) return;
@@ -141,12 +159,17 @@ const useSettingsState = () => {
         async (profileName: string, next: Config, preSharedKey?: string) => {
             const preSharedKeyWrite = preSharedKey === undefined ? {} : { preSharedKey };
             try {
-                await SettingsSvc.SetConfig({
+                const { declined } = await SettingsSvc.SetConfig({
                     ...next,
                     ...preSharedKeyWrite,
                     profileName,
                     username,
                 });
+                // The change needed authorization and the user said no, so the
+                // optimistic update is wrong. Nothing to report: they know.
+                if (declined || reloadOwed.current) {
+                    await reload(profileName);
+                }
             } catch (e) {
                 // The optimistic update is wrong now: the daemon refused it
                 // (a change that needs elevated privileges, an MDM-managed
@@ -206,6 +229,59 @@ const useSettingsState = () => {
         [loaded, save],
     );
 
+    // saveGuardedField applies a setting the daemon restricts to
+    // root/administrator by having the Go side run the app again under the
+    // platform's elevation prompt (UAC, the macOS authentication dialog, polkit).
+    // The prompt is the user's, so the call is made straight from their gesture
+    // and never from the debounce.
+    const saveGuardedField = useCallback(
+        async (k: GuardedField, v: boolean) => {
+            const cur = loadedRef.current;
+            if (!cur) return;
+
+            // Flush what the debounce still owes, before the optimistic update
+            // below joins it: a later save carrying the guarded value would be
+            // refused, and its error dialog would be the second one for a change
+            // the user already authorized.
+            if (saveTimer.current) {
+                clearTimeout(saveTimer.current);
+                saveTimer.current = null;
+                await save(cur.profileName, cur.data);
+            }
+
+            const next: LoadedConfig = {
+                profileName: cur.profileName,
+                data: { ...cur.data, [k]: v },
+            };
+            loadedRef.current = next;
+            setLoaded(next);
+
+            try {
+                await SettingsSvc.SetGuardedSettings({
+                    profileName: cur.profileName,
+                    username,
+                    [k]: v,
+                });
+            } catch (e) {
+                // The daemon is authoritative either way, so re-read before
+                // reporting. A declined prompt is not an error and does not come
+                // through here at all; this is a prompt that could not be raised,
+                // which carries the command that would have done it.
+                await reload(cur.profileName);
+                await errorDialog({
+                    Title: i18next.t("settings.error.saveTitle"),
+                    Message: errorMessage(e),
+                    Command: errorCommand(e),
+                });
+                return;
+            }
+            // Either the change went through or the user declined it. The daemon
+            // says which.
+            await reload(cur.profileName);
+        },
+        [username, save, reload],
+    );
+
     const saveFields = useCallback(
         async (partial: Partial<Config>, opts?: { preSharedKey?: string }) => {
             if (!loaded) return;
@@ -225,15 +301,27 @@ const useSettingsState = () => {
         [loaded, save],
     );
 
-    return { config: loaded?.data ?? null, guiVersion, setField, saveField, saveFields, saveNow };
+    return {
+        config: loaded?.data ?? null,
+        guiVersion,
+        setField,
+        saveField,
+        saveFields,
+        saveGuardedField,
+        saveNow,
+    };
 };
 
 export const SettingsProvider = ({ children }: { children: ReactNode }) => {
-    const { config, guiVersion, setField, saveField, saveFields, saveNow } = useSettingsState();
+    const { config, guiVersion, setField, saveField, saveFields, saveGuardedField, saveNow } =
+        useSettingsState();
 
     const value = useMemo<SettingsContextValue | null>(
-        () => (config ? { config, guiVersion, setField, saveField, saveFields, saveNow } : null),
-        [config, guiVersion, setField, saveField, saveFields, saveNow],
+        () =>
+            config
+                ? { config, guiVersion, setField, saveField, saveFields, saveGuardedField, saveNow }
+                : null,
+        [config, guiVersion, setField, saveField, saveFields, saveGuardedField, saveNow],
     );
 
     if (!value) {

client/ui/frontend/src/hooks/usePrivilege.ts

@@ -1,6 +1,6 @@
 import { useEffect, useState } from "react";
 import { Settings as SettingsSvc } from "@bindings/services";
-import { Privilege } from "@bindings/services/models.js";
+import { type Privilege } from "@bindings/services/models.js";
 
 // usePrivilege reports whether this UI process may perform the changes the daemon
 // restricts to root/administrator. It is answered in-process from our own token

client/ui/frontend/src/lib/connection.ts

@@ -43,7 +43,12 @@ function buildSsoCancelPromise(state: SsoState, signal?: AbortSignal): Promise<v
 }
 
 async function runSsoLogin(
-    result: { verificationUri: string; verificationUriComplete: string; userCode: string },
+    result: {
+        verificationUri: string;
+        verificationUriComplete: string;
+        userCode: string;
+        profileId: string;
+    },
     state: SsoState,
     signal?: AbortSignal,
 ): Promise<void> {
@@ -56,7 +61,7 @@ async function runSsoLogin(
     // suspended, so a frontend-driven Up (a promise continuation) would not
     // fire until the user woke the window (e.g. hovering the tray icon).
     const waitPromise = Connection.WaitSSOLoginAndUp(
-        { userCode: result.userCode, hostname: "" },
+        { userCode: result.userCode, hostname: "", profileId: result.profileId },
         { profileName: "", username: "" },
     );
 

client/ui/frontend/src/modules/session/SessionExpirationDialog.tsx

@@ -11,7 +11,7 @@ import { DialogHeading } from "@/components/dialog/DialogHeading";
 import { SquareIcon } from "@/components/SquareIcon";
 import { Connection, Profiles as ProfilesSvc, Session, WindowManager } from "@bindings/services";
 import { useAutoSizeWindow } from "@/hooks/useAutoSizeWindow";
-import { EVENT_BROWSER_LOGIN_CANCEL } from "@/lib/connection";
+import { EVENT_BROWSER_LOGIN_CANCEL, EVENT_TRIGGER_LOGIN } from "@/lib/connection";
 import { errorDialog, formatErrorMessage } from "@/lib/errors.ts";
 import { formatRemaining } from "@/lib/formatters";
 
@@ -131,6 +131,21 @@ export default function SessionExpirationDialog() {
         }
     }, [busy, t]);
 
+    const authenticate = useCallback(async () => {
+        if (busy) return;
+        setBusy(true);
+        try {
+            await Events.Emit(EVENT_TRIGGER_LOGIN);
+            await WindowManager.CloseSessionExpiration();
+        } catch (e) {
+            setBusy(false);
+            await errorDialog({
+                Title: t("connect.error.loginTitle"),
+                Message: formatErrorMessage(e),
+            });
+        }
+    }, [busy, t]);
+
     const logout = useCallback(async () => {
         if (busy) return;
         setBusy(true);
@@ -185,7 +200,7 @@ export default function SessionExpirationDialog() {
                     variant={"primary"}
                     size={"md"}
                     className={"w-full"}
-                    onClick={stay}
+                    onClick={expired ? authenticate : stay}
                     disabled={busy}
                 >
                     {expired ? t("sessionExpiration.authenticate") : t("sessionExpiration.stay")}

client/ui/frontend/src/modules/settings/SettingsSSH.tsx

@@ -1,3 +1,4 @@
+import { type TFunction } from "i18next";
 import { useTranslation } from "react-i18next";
 import { CopyToClipboard } from "@/components/CopyToClipboard";
 import FancyToggleSwitch from "@/components/switches/FancyToggleSwitch";
@@ -6,51 +7,91 @@ import { Input } from "@/components/inputs/Input";
 import { Label } from "@/components/typography/Label";
 import { cn } from "@/lib/cn";
 import { SectionGroup } from "@/modules/settings/SettingsSection.tsx";
-import { useSettings } from "@/contexts/SettingsContext.tsx";
+import { type GuardedField, useSettings } from "@/contexts/SettingsContext.tsx";
 import { usePrivilege } from "@/hooks/usePrivilege.ts";
-import { Privilege } from "@bindings/services/models.js";
+import type { Privilege } from "@bindings/services/models.js";
 import { type ChangeEvent, type ReactNode, useEffect, useId, useState } from "react";
 
 export function SettingsSSH() {
     const { t } = useTranslation();
-    const { config, setField } = useSettings();
+    const { config, setField, saveGuardedField } = useSettings();
     const privilege = usePrivilege();
+    // The field whose elevation prompt is currently up, if any. The prompt is
+    // modal to the operating system, not to us, so the guarded controls are held
+    // still meanwhile rather than allowed to stack a second one behind it.
+    const [authorizing, setAuthorizing] = useState<GuardedField | null>(null);
     const isSSHServerEnabled = config.serverSshAllowed;
 
+    const authorize = async (field: GuardedField, value: boolean) => {
+        setAuthorizing(field);
+        try {
+            await saveGuardedField(field, value);
+        } finally {
+            setAuthorizing(null);
+        }
+    };
+
     // The daemon restricts only the direction that hands out shells from a process
-    // running as root. So for an unprivileged user a guarded control is either
-    // unavailable (it is off and only they could turn it on) or a one-way switch
-    // (it is on, they may turn it off, but not back on) — say which, either way.
+    // running as root: for all three settings that is switching the field on.
+    //
+    // An unprivileged user gets that direction routed through the platform's
+    // elevation prompt where there is one to raise, and otherwise the old
+    // arrangement, where the control is either unavailable (it is off and only a
+    // privileged caller could turn it on) or a one-way switch (it is on, they may
+    // turn it off but not back on) with the command that does it.
     //
     // A null privilege means we could not determine it: leave the control alone
     // rather than greying it out with nothing to explain why. The daemon enforces
     // this regardless, and a rejected save reports its own guidance.
     const guarded = (
-        guardedDirectionActive: boolean,
+        field: GuardedField,
         command: (p: Privilege) => string,
         // inverted marks a control whose guarded direction is switching it off, so
         // the one-way warning has to read the other way round.
         inverted = false,
     ) => {
+        const plain = (value: boolean) => setField(field, value);
         if (!privilege || privilege.privileged) {
-            return { disabled: false, hint: undefined };
+            return { apply: plain, disabled: false, hint: undefined };
         }
-        const hint = (
-            <PrivilegeHint
-                actor={privilege.actor}
-                command={command(privilege)}
+
+        const guardedDirectionActive = config[field];
+        const hint = (pending: boolean, command?: string) => (
+            <GuardedHint
+                actor={actorLabel(privilege, t)}
                 oneWay={guardedDirectionActive}
                 inverted={inverted}
+                pending={pending}
+                command={command}
             />
         );
-        return { disabled: !guardedDirectionActive, hint };
+
+        if (privilege.canElevate) {
+            return {
+                // Switching off is ours to do; only switching on is authorized.
+                apply: (value: boolean) => {
+                    if (!value) {
+                        plain(value);
+                        return;
+                    }
+                    void authorize(field, value);
+                },
+                disabled: authorizing !== null,
+                hint: hint(authorizing === field),
+            };
+        }
+        return {
+            apply: plain,
+            disabled: !guardedDirectionActive,
+            hint: hint(false, command(privilege)),
+        };
     };
 
-    const sshServer = guarded(config.serverSshAllowed, (p) => p.allowSshServer);
-    const sshRoot = guarded(config.enableSshRoot, (p) => p.enableSshRoot);
+    const sshServer = guarded("serverSshAllowed", (p) => p.allowSshServer);
+    const sshRoot = guarded("enableSshRoot", (p) => p.enableSshRoot);
     // Inverted control: the guarded direction is switching authentication off, so
     // it is the already-disabled state that is the one-way one.
-    const sshAuth = guarded(config.disableSshAuth, (p) => p.disableSshAuth, true);
+    const sshAuth = guarded("disableSshAuth", (p) => p.disableSshAuth, true);
     const jwtTtlId = useId();
     const [jwtTtlInput, setJwtTtlInput] = useState(String(config.sshJwtCacheTtl));
 
@@ -84,7 +125,7 @@ export function SettingsSSH() {
             <SectionGroup title={t("settings.ssh.section.server")}>
                 <FancyToggleSwitch
                     value={config.serverSshAllowed}
-                    onChange={(v) => setField("serverSshAllowed", v)}
+                    onChange={sshServer.apply}
                     disabled={sshServer.disabled}
                     label={t("settings.ssh.server.label")}
                     helpText={t("settings.ssh.server.help")}
@@ -98,7 +139,7 @@ export function SettingsSSH() {
             >
                 <FancyToggleSwitch
                     value={config.enableSshRoot}
-                    onChange={(v) => setField("enableSshRoot", v)}
+                    onChange={sshRoot.apply}
                     disabled={sshRoot.disabled}
                     label={t("settings.ssh.root.label")}
                     helpText={t("settings.ssh.root.help")}
@@ -130,7 +171,7 @@ export function SettingsSSH() {
             >
                 <FancyToggleSwitch
                     value={!config.disableSshAuth}
-                    onChange={(v) => setField("disableSshAuth", !v)}
+                    onChange={(v) => sshAuth.apply(!v)}
                     disabled={sshAuth.disabled}
                     label={t("settings.ssh.jwt.label")}
                     helpText={t("settings.ssh.jwt.help")}
@@ -163,41 +204,81 @@ export function SettingsSSH() {
     );
 }
 
-// PrivilegeHint explains what an unprivileged user can and cannot do with a
-// guarded control, and offers the command that does it with the privileges the
-// daemon requires. oneWay covers the control being in the guarded state already:
-// switching it back is the part that needs privileges.
-function PrivilegeHint({
+// actorLabel names the principal the daemon requires, in the user's language. The
+// Go side reports which one it is rather than wording it, because "administrator
+// privileges" is English and a translated sentence cannot borrow it.
+function actorLabel(privilege: Privilege, t: TFunction): string {
+    return privilege.actorKey === "administrator"
+        ? t("settings.ssh.privilege.actorAdministrator")
+        : t("settings.ssh.privilege.actorRoot");
+}
+
+// GuardedHint is what a control the daemon guards says to an unprivileged user.
+// There are three things worth saying, and it says at most one:
+//
+//   - A prompt is open. Worth a line because it can take a few seconds to appear,
+//     long enough that a control which merely went inert would read as a hang.
+//   - The setting is in its guarded state already (oneWay), so the user may switch
+//     it back as they please and it is switching it away again that will ask. No
+//     command either way: the direction they can take is theirs to take.
+//   - Only a privileged caller can move it at all, and there is no prompt to
+//     raise: the command that does it belongs here, and nothing else will do.
+//
+// Which leaves the case of a control whose guarded direction is still ahead of the
+// user and a prompt that can be raised for it: nothing to say, because clicking it
+// raises the prompt and the prompt explains itself.
+function GuardedHint({
     actor,
-    command,
     oneWay,
     inverted,
+    pending,
+    command,
 }: {
     actor: string;
-    command: string;
     oneWay: boolean;
     inverted: boolean;
+    pending: boolean;
+    command?: string;
 }): ReactNode {
     const { t } = useTranslation();
+
+    if (pending) {
+        return <HintBox>{t("settings.ssh.privilege.authorizePending")}</HintBox>;
+    }
+    if (oneWay) {
+        return (
+            <HintBox>
+                <span>
+                    {inverted
+                        ? t("settings.ssh.privilege.oneWayInverted", { actor })
+                        : t("settings.ssh.privilege.oneWay", { actor })}
+                </span>
+            </HintBox>
+        );
+    }
     if (!command) return null;
+    return (
+        <HintBox>
+            <span>{t("settings.ssh.privilege.hint", { actor })}</span>
+            <CopyToClipboard message={command} alwaysShowIcon wrap variant={"bright"}>
+                <code className={"select-text break-all font-mono text-xs text-nb-gray-200"}>
+                    {command}
+                </code>
+            </CopyToClipboard>
+        </HintBox>
+    );
+}
+
+// HintBox is the box a guarded control puts its explanation in, directly under the
+// control it belongs to.
+function HintBox({ children }: { children: ReactNode }): ReactNode {
     return (
         <div
             className={
                 "-mt-2 flex flex-col gap-1 rounded-md bg-nb-gray-930 px-3 py-2 text-xs text-nb-gray-300"
             }
         >
-            <span>
-                {!oneWay
-                    ? t("settings.ssh.privilege.hint", { actor })
-                    : inverted
-                      ? t("settings.ssh.privilege.oneWayInverted", { actor })
-                      : t("settings.ssh.privilege.oneWay", { actor })}
-            </span>
-            <CopyToClipboard message={command} alwaysShowIcon wrap variant={"bright"}>
-                <code className={"select-text break-all font-mono text-xs text-nb-gray-200"}>
-                    {command}
-                </code>
-            </CopyToClipboard>
+            {children}
         </div>
     );
 }

client/ui/i18n/locales/de/common.json

@@ -401,9 +401,6 @@
     "networks.bulk.label": {
         "message": "Alle sichtbaren Ressourcen umschalten"
     },
-    "settings.nav.label": {
-        "message": "Einstellungsbereiche"
-    },
     "profile.switch.title": {
         "message": "Zu Profil \"{name}\" wechseln?"
     },
@@ -497,6 +494,9 @@
     "settings.error.debugBundleTitle": {
         "message": "Debug-Paket fehlgeschlagen"
     },
+    "settings.nav.label": {
+        "message": "Einstellungsbereiche"
+    },
     "settings.tabs.general": {
         "message": "Allgemein"
     },
@@ -1330,5 +1330,29 @@
     },
     "error.unknown": {
         "message": "Vorgang fehlgeschlagen."
+    },
+    "error.elevation_unavailable": {
+        "message": "NetBird konnte auf diesem System nicht die nötigen Rechte anfordern. Führen Sie stattdessen dies aus:"
+    },
+    "error.elevation_failed": {
+        "message": "Die Änderung konnte mit erhöhten Rechten nicht angewendet werden. Führen Sie stattdessen dies aus:"
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "root-Rechte"
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "Administratorrechte"
+    },
+    "settings.ssh.privilege.hint": {
+        "message": "Erfordert {actor}. Führen Sie stattdessen dies aus:"
+    },
+    "settings.ssh.privilege.oneWay": {
+        "message": "Sie können dies deaktivieren, zum erneuten Aktivieren sind {actor} erforderlich."
+    },
+    "settings.ssh.privilege.oneWayInverted": {
+        "message": "Sie können dies aktivieren, zum erneuten Deaktivieren sind {actor} erforderlich."
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "Warten auf Autorisierung…"
     }
 }

client/ui/i18n/locales/en/common.json

@@ -1775,16 +1775,36 @@
         "message": "Operation failed.",
         "description": "Generic fallback error message used when no specific error applies."
     },
+    "error.elevation_unavailable": {
+        "message": "NetBird could not ask this system for the privileges the change needs. Run this instead:",
+        "description": "Error: this computer has no way to prompt for elevated privileges. Followed by a copyable command that applies the setting from a terminal."
+    },
+    "error.elevation_failed": {
+        "message": "The change could not be applied with elevated privileges. Run this instead:",
+        "description": "Error: the authorization succeeded but applying the setting afterwards failed. Followed by a copyable command that applies the setting from a terminal."
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "root",
+        "description": "Fills {actor} in the settings.ssh.privilege.* messages on Linux, macOS and BSD, where the daemon requires the root account. 'root' is an account name and stays as it is; add the word for privileges or rights around it if the sentence needs one to read naturally."
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "administrator privileges",
+        "description": "Fills {actor} in the settings.ssh.privilege.* messages on Windows, where the daemon requires an elevated administrator. The Windows term for the rights an account is asked to elevate to."
+    },
     "settings.ssh.privilege.hint": {
         "message": "Requires {actor}. Run this instead:",
         "description": "Help text under an SSH setting the user cannot change: it needs elevated privileges. {actor} is 'root' on Linux/macOS or 'administrator privileges' on Windows. Followed by a copyable command."
     },
     "settings.ssh.privilege.oneWay": {
-        "message": "You can switch this off, but switching it back on needs {actor}:",
-        "description": "Warning under an SSH setting an unprivileged user may disable but not re-enable. {actor} is 'root' on Linux/macOS or 'administrator privileges' on Windows. Followed by a copyable command."
+        "message": "You can switch this off, but switching it back on needs {actor}.",
+        "description": "Help text under an SSH setting that is already on: an unprivileged user may switch it off freely, and switching it on again is what needs the privileges. No command follows, since the direction they can take is theirs to take. {actor} is 'root' on Linux/macOS or 'administrator privileges' on Windows."
     },
     "settings.ssh.privilege.oneWayInverted": {
-        "message": "You can switch this on, but switching it back off needs {actor}:",
-        "description": "Warning under the SSH authentication setting, which an unprivileged user may re-enable but not disable again. {actor} is 'root' on Linux/macOS or 'administrator privileges' on Windows. Followed by a copyable command."
+        "message": "You can switch this on, but switching it back off needs {actor}.",
+        "description": "Same as settings.ssh.privilege.oneWay, for the SSH authentication setting once it has been switched off: switching it off again is what needs the privileges."
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "Waiting for authorization…",
+        "description": "Replaces the help text under a guarded SSH setting while the authorization prompt is open, which can take a few seconds to appear. Keep the trailing ellipsis."
     }
 }

client/ui/i18n/locales/es/common.json

@@ -401,9 +401,6 @@
     "networks.bulk.label": {
         "message": "Conmutar todos los recursos visibles"
     },
-    "settings.nav.label": {
-        "message": "Secciones de configuración"
-    },
     "profile.switch.title": {
         "message": "¿Cambiar el perfil a «{name}»?"
     },
@@ -497,6 +494,9 @@
     "settings.error.debugBundleTitle": {
         "message": "Error en el paquete de diagnóstico"
     },
+    "settings.nav.label": {
+        "message": "Secciones de configuración"
+    },
     "settings.tabs.general": {
         "message": "General"
     },
@@ -1330,5 +1330,29 @@
     },
     "error.unknown": {
         "message": "La operación falló."
+    },
+    "error.elevation_unavailable": {
+        "message": "NetBird no pudo solicitar a este sistema los privilegios necesarios. Ejecute esto en su lugar:"
+    },
+    "error.elevation_failed": {
+        "message": "No se pudo aplicar el cambio con privilegios elevados. Ejecute esto en su lugar:"
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "privilegios de root"
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "privilegios de administrador"
+    },
+    "settings.ssh.privilege.hint": {
+        "message": "Requiere {actor}. Ejecute esto en su lugar:"
+    },
+    "settings.ssh.privilege.oneWay": {
+        "message": "Puede desactivarlo, pero volver a activarlo requiere {actor}."
+    },
+    "settings.ssh.privilege.oneWayInverted": {
+        "message": "Puede activarlo, pero volver a desactivarlo requiere {actor}."
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "Esperando la autorización…"
     }
 }

client/ui/i18n/locales/fr/common.json

@@ -401,9 +401,6 @@
     "networks.bulk.label": {
         "message": "Activer/désactiver toutes les ressources visibles"
     },
-    "settings.nav.label": {
-        "message": "Sections des paramètres"
-    },
     "profile.switch.title": {
         "message": "Basculer vers le profil « {name} » ?"
     },
@@ -497,6 +494,9 @@
     "settings.error.debugBundleTitle": {
         "message": "Échec du lot de diagnostic"
     },
+    "settings.nav.label": {
+        "message": "Sections des paramètres"
+    },
     "settings.tabs.general": {
         "message": "Général"
     },
@@ -1330,5 +1330,29 @@
     },
     "error.unknown": {
         "message": "L’opération a échoué."
+    },
+    "error.elevation_unavailable": {
+        "message": "NetBird n’a pas pu demander à ce système les privilèges nécessaires. Exécutez plutôt ceci :"
+    },
+    "error.elevation_failed": {
+        "message": "La modification n’a pas pu être appliquée avec des privilèges élevés. Exécutez plutôt ceci :"
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "les privilèges root"
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "les privilèges administrateur"
+    },
+    "settings.ssh.privilege.hint": {
+        "message": "Nécessite {actor}. Exécutez plutôt ceci :"
+    },
+    "settings.ssh.privilege.oneWay": {
+        "message": "Vous pouvez le désactiver, mais le réactiver nécessite {actor}."
+    },
+    "settings.ssh.privilege.oneWayInverted": {
+        "message": "Vous pouvez l’activer, mais le désactiver de nouveau nécessite {actor}."
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "En attente de l’autorisation…"
     }
 }

client/ui/i18n/locales/hu/common.json

@@ -401,9 +401,6 @@
     "networks.bulk.label": {
         "message": "Összes látható erőforrás be/ki"
     },
-    "settings.nav.label": {
-        "message": "Beállítások szakaszai"
-    },
     "profile.switch.title": {
         "message": "Váltás a(z) \"{name}\" profilra?"
     },
@@ -497,6 +494,9 @@
     "settings.error.debugBundleTitle": {
         "message": "Hibakeresési csomag sikertelen"
     },
+    "settings.nav.label": {
+        "message": "Beállítások szakaszai"
+    },
     "settings.tabs.general": {
         "message": "Általános"
     },
@@ -1330,5 +1330,29 @@
     },
     "error.unknown": {
         "message": "A művelet meghiúsult."
+    },
+    "error.elevation_unavailable": {
+        "message": "A NetBird nem tudta bekérni a rendszertől a szükséges jogosultságokat. Futtassa inkább ezt:"
+    },
+    "error.elevation_failed": {
+        "message": "A módosítást emelt szintű jogosultságokkal sem sikerült alkalmazni. Futtassa inkább ezt:"
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "root jogosultság"
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "rendszergazdai jogosultság"
+    },
+    "settings.ssh.privilege.hint": {
+        "message": "{actor} szükséges hozzá. Futtassa inkább ezt:"
+    },
+    "settings.ssh.privilege.oneWay": {
+        "message": "Kikapcsolhatja, de a visszakapcsolásához {actor} szükséges."
+    },
+    "settings.ssh.privilege.oneWayInverted": {
+        "message": "Bekapcsolhatja, de az ismételt kikapcsolásához {actor} szükséges."
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "Várakozás az engedélyezésre…"
     }
 }

client/ui/i18n/locales/it/common.json

@@ -401,9 +401,6 @@
     "networks.bulk.label": {
         "message": "Attiva/disattiva tutte le risorse visibili"
     },
-    "settings.nav.label": {
-        "message": "Sezioni delle impostazioni"
-    },
     "profile.switch.title": {
         "message": "Passare al profilo «{name}»?"
     },
@@ -497,6 +494,9 @@
     "settings.error.debugBundleTitle": {
         "message": "Pacchetto di debug non riuscito"
     },
+    "settings.nav.label": {
+        "message": "Sezioni delle impostazioni"
+    },
     "settings.tabs.general": {
         "message": "Generale"
     },
@@ -1330,5 +1330,29 @@
     },
     "error.unknown": {
         "message": "Operazione non riuscita."
+    },
+    "error.elevation_unavailable": {
+        "message": "NetBird non ha potuto richiedere a questo sistema i privilegi necessari. Esegua invece questo:"
+    },
+    "error.elevation_failed": {
+        "message": "Non è stato possibile applicare la modifica con privilegi elevati. Esegua invece questo:"
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "i privilegi di root"
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "i privilegi di amministratore"
+    },
+    "settings.ssh.privilege.hint": {
+        "message": "Richiede {actor}. Esegua invece questo:"
+    },
+    "settings.ssh.privilege.oneWay": {
+        "message": "Può disabilitarlo, ma riabilitarlo richiede {actor}."
+    },
+    "settings.ssh.privilege.oneWayInverted": {
+        "message": "Può abilitarlo, ma disabilitarlo di nuovo richiede {actor}."
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "In attesa dell'autorizzazione…"
     }
 }

client/ui/i18n/locales/ja/common.json

@@ -1304,6 +1304,9 @@
     "daemon.outdated.description": {
         "message": "このアプリを使用するには NetBird サービスを更新してください。"
     },
+    "daemon.outdated.download": {
+        "message": "最新版をダウンロード"
+    },
     "error.jwt_clock_skew": {
         "message": "サインインに失敗しました: このデバイスの時計がサーバーと同期していません。システムの時計を同期してからもう一度お試しください。"
     },
@@ -1327,5 +1330,29 @@
     },
     "error.unknown": {
         "message": "操作に失敗しました。"
+    },
+    "error.elevation_unavailable": {
+        "message": "NetBird はこのシステムに必要な権限を要求できませんでした。代わりに次のコマンドを実行してください:"
+    },
+    "error.elevation_failed": {
+        "message": "昇格した権限でも変更を適用できませんでした。代わりに次のコマンドを実行してください:"
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "root 権限"
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "管理者権限"
+    },
+    "settings.ssh.privilege.hint": {
+        "message": "{actor}が必要です。代わりに次のコマンドを実行してください:"
+    },
+    "settings.ssh.privilege.oneWay": {
+        "message": "無効にはできますが、再度有効にするには{actor}が必要です。"
+    },
+    "settings.ssh.privilege.oneWayInverted": {
+        "message": "有効にはできますが、再度無効にするには{actor}が必要です。"
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "承認を待っています…"
     }
 }

client/ui/i18n/locales/pt/common.json

@@ -401,9 +401,6 @@
     "networks.bulk.label": {
         "message": "Alternar todos os recursos visíveis"
     },
-    "settings.nav.label": {
-        "message": "Seções das configurações"
-    },
     "profile.switch.title": {
         "message": "Alternar perfil para \"{name}\"?"
     },
@@ -497,6 +494,9 @@
     "settings.error.debugBundleTitle": {
         "message": "Falha no pacote de depuração"
     },
+    "settings.nav.label": {
+        "message": "Seções das configurações"
+    },
     "settings.tabs.general": {
         "message": "Geral"
     },
@@ -1330,5 +1330,29 @@
     },
     "error.unknown": {
         "message": "A operação falhou."
+    },
+    "error.elevation_unavailable": {
+        "message": "O NetBird não conseguiu solicitar a este sistema os privilégios necessários. Execute isto em vez disso:"
+    },
+    "error.elevation_failed": {
+        "message": "Não foi possível aplicar a alteração com privilégios elevados. Execute isto em vez disso:"
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "privilégios de root"
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "privilégios de administrador"
+    },
+    "settings.ssh.privilege.hint": {
+        "message": "Requer {actor}. Execute isto em vez disso:"
+    },
+    "settings.ssh.privilege.oneWay": {
+        "message": "Você pode desativar isto, mas ativar novamente requer {actor}."
+    },
+    "settings.ssh.privilege.oneWayInverted": {
+        "message": "Você pode ativar isto, mas desativar novamente requer {actor}."
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "Aguardando a autorização…"
     }
 }

client/ui/i18n/locales/ru/common.json

@@ -401,9 +401,6 @@
     "networks.bulk.label": {
         "message": "Переключить все видимые ресурсы"
     },
-    "settings.nav.label": {
-        "message": "Разделы настроек"
-    },
     "profile.switch.title": {
         "message": "Переключиться на профиль «{name}»?"
     },
@@ -497,6 +494,9 @@
     "settings.error.debugBundleTitle": {
         "message": "Не удалось создать отладочный пакет"
     },
+    "settings.nav.label": {
+        "message": "Разделы настроек"
+    },
     "settings.tabs.general": {
         "message": "Общие"
     },
@@ -1330,5 +1330,29 @@
     },
     "error.unknown": {
         "message": "Не удалось выполнить операцию."
+    },
+    "error.elevation_unavailable": {
+        "message": "NetBird не смог запросить у этой системы нужные права. Выполните вместо этого:"
+    },
+    "error.elevation_failed": {
+        "message": "Не удалось применить изменение с повышенными правами. Выполните вместо этого:"
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "права root"
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "права администратора"
+    },
+    "settings.ssh.privilege.hint": {
+        "message": "Требуются {actor}. Выполните вместо этого:"
+    },
+    "settings.ssh.privilege.oneWay": {
+        "message": "Отключить можно, но чтобы включить снова, нужны {actor}."
+    },
+    "settings.ssh.privilege.oneWayInverted": {
+        "message": "Включить можно, но чтобы отключить снова, нужны {actor}."
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "Ожидание авторизации…"
     }
 }

client/ui/i18n/locales/zh-CN/common.json

@@ -401,9 +401,6 @@
     "networks.bulk.label": {
         "message": "切换所有可见资源"
     },
-    "settings.nav.label": {
-        "message": "设置部分"
-    },
     "profile.switch.title": {
         "message": "切换到配置文件“{name}”？"
     },
@@ -497,6 +494,9 @@
     "settings.error.debugBundleTitle": {
         "message": "创建调试包失败"
     },
+    "settings.nav.label": {
+        "message": "设置部分"
+    },
     "settings.tabs.general": {
         "message": "常规"
     },
@@ -1330,5 +1330,29 @@
     },
     "error.unknown": {
         "message": "操作失败。"
+    },
+    "error.elevation_unavailable": {
+        "message": "NetBird 无法向此系统请求所需的权限。请改为运行："
+    },
+    "error.elevation_failed": {
+        "message": "即使使用提升的权限也无法应用此更改。请改为运行："
+    },
+    "settings.ssh.privilege.actorRoot": {
+        "message": "root 权限"
+    },
+    "settings.ssh.privilege.actorAdministrator": {
+        "message": "管理员权限"
+    },
+    "settings.ssh.privilege.hint": {
+        "message": "需要{actor}。请改为运行："
+    },
+    "settings.ssh.privilege.oneWay": {
+        "message": "您可以关闭此项，但重新开启需要{actor}。"
+    },
+    "settings.ssh.privilege.oneWayInverted": {
+        "message": "您可以开启此项，但再次关闭需要{actor}。"
+    },
+    "settings.ssh.privilege.authorizePending": {
+        "message": "正在等待授权…"
     }
 }

client/ui/main.go

@@ -8,13 +8,13 @@ import (
 	"flag"
 	"io/fs"
 	"log"
+	"os"
 	"runtime"
 	"strings"
 
 	"github.com/sirupsen/logrus"
 	"github.com/wailsapp/wails/v3/pkg/application"
 	"github.com/wailsapp/wails/v3/pkg/events"
-	"github.com/wailsapp/wails/v3/pkg/services/notifications"
 
 	"github.com/netbirdio/netbird/client/ui/authsession"
 	"github.com/netbirdio/netbird/client/ui/i18n"
@@ -63,7 +63,7 @@ type registeredServices struct {
 	profiles        *services.Profiles
 	update          *services.Update
 	daemonFeed      *services.DaemonFeed
-	notifier        *notifications.NotificationService
+	notifier        *Notifier
 	compat          *services.Compat
 	profileSwitcher *services.ProfileSwitcher
 	bundle          *i18n.Bundle
@@ -80,6 +80,14 @@ func init() {
 }
 
 func main() {
+	// The one-shot that applies the settings the daemon restricts to
+	// root/administrator, which this binary runs itself as under the platform's
+	// elevation prompt. Handled before anything GUI so no window, tray or
+	// single-instance lock is involved.
+	if services.IsPrivilegedSettingsRun(os.Args[1:]) {
+		os.Exit(runPrivilegedSettings(os.Args[1:]))
+	}
+
 	daemonAddr, userSetLogFile := parseFlagsAndInitLog()
 	conn := NewConn(daemonAddr)
 
@@ -102,7 +110,7 @@ func main() {
 	updaterHolder := updater.NewHolder(app.Event)
 	update := services.NewUpdate(conn, updaterHolder)
 	daemonFeed := services.NewDaemonFeed(conn, app.Event, updaterHolder, debugLog)
-	notifier := notifications.New()
+	notifier := newNotifier()
 	compat := services.NewCompat(conn)
 	// macOS shows no toast until permission is requested. Run it after
 	// ApplicationStarted so the notifier's Startup has initialised the
@@ -210,7 +218,7 @@ func main() {
 // requestNotificationAuthorization prompts for macOS notification permission.
 // The request blocks until the user responds (up to 3 minutes), so callers run
 // it in a goroutine. No-op on Linux/Windows.
-func requestNotificationAuthorization(notifier *notifications.NotificationService) {
+func requestNotificationAuthorization(notifier *Notifier) {
 	authorized, err := notifier.CheckNotificationAuthorization()
 	if err != nil {
 		logrus.Debugf("check notification authorization: %v", err)

client/ui/notifier.go

@@ -0,0 +1,101 @@
+//go:build !android && !ios && !freebsd && !js
+
+package main
+
+import (
+	"context"
+	"errors"
+	"sync/atomic"
+
+	log "github.com/sirupsen/logrus"
+	"github.com/wailsapp/wails/v3/pkg/application"
+	"github.com/wailsapp/wails/v3/pkg/services/notifications"
+)
+
+var errNotificationsUnavailable = errors.New("notifications unavailable")
+
+// Notifier wraps the Wails notification service so an unavailable backend
+// disables notifications instead of aborting the app. Startup fails for
+// environment reasons (a bare unbundled binary on macOS has no bundle
+// identifier, a headless Linux session has no D-Bus session bus), and Wails
+// treats a service startup error as fatal. After a failed startup every call
+// is a no-op: on macOS, touching UNUserNotificationCenter without a bundle
+// identifier raises an Objective-C exception that recover() cannot catch.
+type Notifier struct {
+	inner     *notifications.NotificationService
+	available atomic.Bool
+}
+
+func newNotifier() *Notifier {
+	return &Notifier{inner: notifications.New()}
+}
+
+// ServiceName implements the Wails service-name hook for startup logs.
+func (n *Notifier) ServiceName() string {
+	return n.inner.ServiceName()
+}
+
+// ServiceStartup starts the platform notifier, downgrading failure to a
+// warning so the app keeps running without notifications.
+func (n *Notifier) ServiceStartup(ctx context.Context, options application.ServiceOptions) error {
+	if err := n.inner.ServiceStartup(ctx, options); err != nil {
+		log.Warnf("notifications disabled: %v", err)
+		return nil
+	}
+	n.available.Store(true)
+	return nil
+}
+
+func (n *Notifier) ServiceShutdown() error {
+	if !n.available.Load() {
+		return nil
+	}
+	return n.inner.ServiceShutdown()
+}
+
+func (n *Notifier) CheckNotificationAuthorization() (bool, error) {
+	if !n.available.Load() {
+		return false, errNotificationsUnavailable
+	}
+	return n.inner.CheckNotificationAuthorization()
+}
+
+func (n *Notifier) RequestNotificationAuthorization() (bool, error) {
+	if !n.available.Load() {
+		return false, errNotificationsUnavailable
+	}
+	return n.inner.RequestNotificationAuthorization()
+}
+
+// SendNotification delivers a notification, silently dropping it when the
+// backend never started (notifications are best-effort everywhere).
+func (n *Notifier) SendNotification(options notifications.NotificationOptions) error {
+	if !n.available.Load() {
+		log.Debugf("notifications disabled, dropping %q", options.ID)
+		return nil
+	}
+	return n.inner.SendNotification(options)
+}
+
+func (n *Notifier) SendNotificationWithActions(options notifications.NotificationOptions) error {
+	if !n.available.Load() {
+		log.Debugf("notifications disabled, dropping %q", options.ID)
+		return nil
+	}
+	return n.inner.SendNotificationWithActions(options)
+}
+
+func (n *Notifier) RegisterNotificationCategory(category notifications.NotificationCategory) error {
+	if !n.available.Load() {
+		return nil
+	}
+	return n.inner.RegisterNotificationCategory(category)
+}
+
+// OnNotificationResponse registers the response callback. Pure Go state, so
+// it is safe (and simply inert) when the backend never started.
+//
+//wails:ignore
+func (n *Notifier) OnNotificationResponse(callback func(result notifications.NotificationResult)) {
+	n.inner.OnNotificationResponse(callback)
+}

client/ui/preferences/store.go

@@ -246,6 +246,7 @@ func (s *Store) ExistedAtLoad() bool {
 func (s *Store) load() error {
 	if _, err := os.Stat(s.path); err != nil {
 		if errors.Is(err, os.ErrNotExist) {
+			log.Infof("no ui preferences file at %s; using defaults", s.path)
 			return nil
 		}
 		return fmt.Errorf("stat preferences: %w", err)

client/ui/privileged_settings.go

@@ -0,0 +1,27 @@
+//go:build !android && !ios && !freebsd && !js
+
+package main
+
+import (
+	"github.com/netbirdio/netbird/client/proto"
+	"github.com/netbirdio/netbird/client/ui/services"
+)
+
+// The one-shot mode this binary runs itself in, elevated, to apply the settings the
+// daemon restricts to root/administrator. It is handled before anything GUI, so no
+// window, tray or single-instance lock is involved.
+//
+// Only the wiring is here: what the mode accepts and does lives beside the code
+// that asks for it, in services.RunPrivilegedSettings, so the settings it will
+// apply are declared once. There is nothing privileged about the mode itself; it
+// sends the same request the frontend would have sent, and the daemon authorizes it
+// from the identity the kernel reports on the control channel exactly as it does
+// for `sudo netbird up`.
+func runPrivilegedSettings(args []string) int {
+	return services.RunPrivilegedSettings(args, func(addr string) (proto.DaemonServiceClient, error) {
+		if addr == "" {
+			addr = DaemonAddr()
+		}
+		return NewConn(addr).Client()
+	})
+}

client/ui/services/connection.go

@@ -33,12 +33,21 @@ type LoginResult struct {
 	UserCode                string `json:"userCode"`
 	VerificationURI         string `json:"verificationUri"`
 	VerificationURIComplete string `json:"verificationUriComplete"`
+	// ProfileID is the ID of the profile this login ran against, or "" when the
+	// caller named the profile itself and no ID was resolved. Pass it back in
+	// WaitSSOParams so the account email lands on this profile even if the
+	// active one changes during SSO.
+	ProfileID string `json:"profileId"`
 }
 
 // WaitSSOParams are the inputs to waitSSOLogin.
 type WaitSSOParams struct {
 	UserCode string `json:"userCode"`
 	Hostname string `json:"hostname"`
+	// ProfileID is the profile the login was started for, used to file the
+	// account email against it rather than against whichever profile is active
+	// when the flow returns. Optional: empty falls back to the active profile.
+	ProfileID string `json:"profileId"`
 }
 
 // UpParams selects the profile to bring up.
@@ -77,11 +86,16 @@ func (s *Connection) Login(ctx context.Context, p LoginParams) (LoginResult, err
 	// Fall back to the daemon's active profile and the current OS user.
 	profileName := p.ProfileName
 	username := p.Username
+	// Only set when the daemon told us the ID. A caller-supplied ProfileName is
+	// a handle — a display name or an ID prefix resolve too — and the state file
+	// is named after the ID, so passing a handle on would name the wrong file.
+	profileID := ""
 	if profileName == "" {
 		if active, aerr := cli.GetActiveProfile(ctx, &proto.GetActiveProfileRequest{}); aerr == nil {
 			// Address the active profile by ID (the daemon resolves it as a
 			// handle); names can collide, the ID cannot.
 			profileName = active.GetId()
+			profileID = profileName
 			if username == "" {
 				username = active.GetUsername()
 			}
@@ -122,6 +136,7 @@ func (s *Connection) Login(ctx context.Context, p LoginParams) (LoginResult, err
 		UserCode:                resp.GetUserCode(),
 		VerificationURI:         resp.GetVerificationURI(),
 		VerificationURIComplete: resp.GetVerificationURIComplete(),
+		ProfileID:               profileID,
 	}, nil
 }
 
@@ -242,6 +257,31 @@ func (s *Connection) waitSSOLogin(ctx context.Context, p WaitSSOParams) (string,
 		return "", s.classifyDaemonError(err)
 	}
 	log.Infof("SSO login completed, daemon reported success")
+
+	// Persist the account email the same way the CLI does after its own
+	// WaitSSOLogin: the daemon returns it but cannot store it, since it runs as
+	// root and the per-profile state file is user-owned (see Logout below).
+	// Without this the profile has no email, so Profiles.List shows no account
+	// and later logins and session extends go out without a login_hint —
+	// leaving the IdP to guess which account was meant.
+	if email := resp.GetEmail(); email != "" {
+		state := &profilemanager.ProfileState{Email: email}
+		pm := profilemanager.NewProfileManager()
+
+		// Against the profile the login was started for: SSO spans seconds of
+		// user interaction, and a profile switch in that window would otherwise
+		// file the email under the wrong profile.
+		if p.ProfileID != "" {
+			err = pm.SetProfileState(profilemanager.ID(p.ProfileID), state)
+		} else {
+			err = pm.SetActiveProfileState(state)
+		}
+		if err != nil {
+			// Non-fatal: the login itself succeeded.
+			log.Warnf("failed to store account email: %v", err)
+		}
+	}
+
 	return resp.GetEmail(), nil
 }
 

client/ui/services/guarded.go

@@ -0,0 +1,226 @@
+//go:build !android && !ios && !freebsd && !js
+
+package services
+
+import (
+	"context"
+	"errors"
+	"fmt"
+	"strings"
+	"time"
+
+	log "github.com/sirupsen/logrus"
+
+	"github.com/netbirdio/netbird/client/internal/elevate"
+	"github.com/netbirdio/netbird/client/internal/ipcauth"
+)
+
+// The command line of the one-shot mode this binary runs itself in, elevated, to
+// apply a setting the daemon restricts to root/administrator. The setting flags
+// spell the same words as `netbird up`, so the command a user is shown and what
+// runs behind the prompt read alike. Parsed in oneshot.go.
+const (
+	FlagApplyPrivilegedSettings = "apply-privileged-settings"
+	FlagDaemonAddr              = "daemon-addr"
+	FlagProfile                 = "profile"
+	FlagUser                    = "user"
+	FlagLogLevel                = "log-level"
+	FlagManagementURL           = "management-url"
+	FlagAllowServerSSH          = "allow-server-ssh"
+	FlagEnableSSHRoot           = "enable-ssh-root"
+	FlagDisableSSHAuth          = "disable-ssh-auth"
+)
+
+// Error codes for the ways asking for privileges can fail.
+const (
+	CodeElevationUnavailable = "elevation_unavailable"
+	CodeElevationFailed      = "elevation_failed"
+)
+
+// elevationTimeout bounds the wait for a prompt and the change behind it, so a
+// dialog nobody answers does not leave its control disabled for the session. Long
+// enough to find a password manager, and no shorter than the platforms' own prompt
+// timeouts: Windows gives up on its consent dialog after two minutes by itself.
+//
+// It always ends our waiting, and not always the prompt: Security.framework offers
+// no way to withdraw a request, so on macOS the system's own timeout is what closes
+// the dialog.
+const elevationTimeout = 5 * time.Minute
+
+// elevator raises the platform's privilege prompt and runs the change behind it.
+// An interface so tests can answer without a prompt.
+type elevator interface {
+	// Run runs this binary again, elevated, with the given arguments.
+	Run(ctx context.Context, args ...string) error
+	// Available reports whether there is a prompt to raise on this host at all.
+	Available() bool
+}
+
+// osElevator is the real thing: see the elevate package.
+type osElevator struct{}
+
+func (osElevator) Run(ctx context.Context, args ...string) error {
+	return elevate.Run(ctx, args...)
+}
+
+func (osElevator) Available() bool {
+	return elevate.Available()
+}
+
+// SaveOutcome reports what became of a change that needed authorization.
+//
+// A declined prompt is a result, not an error: the user was asked and said no, so
+// nothing was applied and nothing went wrong. Reporting it as an error would have
+// every cancelled prompt logged as one.
+type SaveOutcome struct {
+	// Declined is set when the user dismissed the authorization prompt, or was
+	// refused by policy. Nothing was changed.
+	Declined bool `json:"declined"`
+}
+
+// GuardedSettings is the subset of the config the daemon restricts to
+// root/administrator. Only the fields that are set are changed: a nil pointer, or
+// an empty management URL, leaves that setting alone.
+//
+// The management URL is in here because pointing a host with the SSH server
+// running at another management identity hands the decision of who may open a
+// shell on it to whoever runs that server, which is the same power as enabling
+// the SSH server in the first place.
+type GuardedSettings struct {
+	ProfileName      string `json:"profileName"`
+	Username         string `json:"username"`
+	ManagementURL    string `json:"managementUrl,omitempty"`
+	ServerSSHAllowed *bool  `json:"serverSshAllowed,omitempty"`
+	EnableSSHRoot    *bool  `json:"enableSshRoot,omitempty"`
+	DisableSSHAuth   *bool  `json:"disableSshAuth,omitempty"`
+}
+
+// guardedSetting is one setting to change, in the two spellings this needs: the
+// one-shot's own flag, and the `netbird up` flag that does the same thing from a
+// terminal, for when there is no prompt to raise.
+type guardedSetting struct {
+	arg  string
+	flag string
+}
+
+// SetGuardedSettings applies settings the daemon refuses from an unprivileged
+// caller, by having the operating system run this binary again, elevated, to send
+// the same request the frontend would have sent itself.
+//
+// The user authorizes it at the platform's own prompt: the UAC consent dialog,
+// the macOS authentication dialog, or the polkit agent's. Any credentials are the
+// operating system's business; NetBird neither sees nor asks for them. Nothing
+// about the daemon's rules changes, and the elevated process is authorized like
+// any other privileged caller, from the identity the kernel reports for it.
+//
+// A declined prompt comes back as SaveOutcome.Declined with no error. When there is
+// no prompt to raise, or the elevated run failed, the error carries the command
+// that does the same thing from a terminal.
+func (s *Settings) SetGuardedSettings(ctx context.Context, p GuardedSettings) (SaveOutcome, error) {
+	settings := guardedSettings(p)
+	if len(settings) == 0 {
+		return SaveOutcome{}, &ClientError{
+			Code:  CodeElevationFailed,
+			Short: "no setting to apply",
+			Long:  "no setting to apply",
+		}
+	}
+
+	args := append([]string{
+		"--" + FlagApplyPrivilegedSettings,
+		"--" + FlagDaemonAddr, s.daemonAddr,
+		"--" + FlagProfile, p.ProfileName,
+		"--" + FlagUser, p.Username,
+	}, oneShotArgs(settings)...)
+
+	ctx, cancel := context.WithTimeout(ctx, elevationTimeout)
+	defer cancel()
+
+	// These changes hand out shells on this host, so both ends are logged: when the
+	// prompt went up, and what came of it. It is also the only account of a prompt
+	// that was slow to appear or never answered.
+	log.Infof("asking for privileges to apply %s", guardedSummary(p))
+
+	if err := s.elevator.Run(ctx, args...); err != nil {
+		return s.elevationOutcome(err, p)
+	}
+
+	log.Infof("applied %s with the privileges the user authorized", guardedSummary(p))
+	return SaveOutcome{}, nil
+}
+
+// elevationOutcome sorts what came back into the one normal ending and the two
+// that need reporting, with the command that does the same thing by hand.
+func (s *Settings) elevationOutcome(err error, p GuardedSettings) (SaveOutcome, error) {
+	switch {
+	case errors.Is(err, elevate.ErrDeclined):
+		// With the reason: an account that may not elevate at all lands here too,
+		// and the log is the only place that says which it was.
+		log.Infof("the elevation prompt for %s was declined: %v", guardedSummary(p), err)
+		return SaveOutcome{Declined: true}, nil
+	case errors.Is(err, elevate.ErrUnavailable):
+		log.Warnf("cannot ask for privileges to apply %s: %v", guardedSummary(p), err)
+		return SaveOutcome{}, &ClientError{
+			Code:    CodeElevationUnavailable,
+			Short:   s.classifier.translateShort(CodeElevationUnavailable),
+			Long:    err.Error(),
+			Command: guardedCommand(p),
+		}
+	default:
+		log.Errorf("applying %s with elevated privileges failed: %v", guardedSummary(p), err)
+		return SaveOutcome{}, &ClientError{
+			Code:    CodeElevationFailed,
+			Short:   s.classifier.translateShort(CodeElevationFailed),
+			Long:    err.Error(),
+			Command: guardedCommand(p),
+		}
+	}
+}
+
+// guardedSettings renders the settings that are actually being changed, from the
+// same table the one-shot parses them with: see oneshot.go.
+func guardedSettings(p GuardedSettings) []guardedSetting {
+	var settings []guardedSetting
+	for _, field := range guardedFields {
+		value, ok := field.read(p)
+		if !ok {
+			continue
+		}
+		settings = append(settings, guardedSetting{
+			arg:  "--" + field.flag + "=" + value,
+			flag: field.up(value),
+		})
+	}
+	return settings
+}
+
+func oneShotArgs(settings []guardedSetting) []string {
+	args := make([]string, 0, len(settings))
+	for _, setting := range settings {
+		args = append(args, setting.arg)
+	}
+	return args
+}
+
+func upFlags(settings []guardedSetting) []string {
+	flags := make([]string, 0, len(settings))
+	for _, setting := range settings {
+		flags = append(flags, setting.flag)
+	}
+	return flags
+}
+
+// guardedCommand is the elevated command line equivalent to the requested
+// change, the same shape the daemon names in its own refusals.
+func guardedCommand(p GuardedSettings) string {
+	settings := guardedSettings(p)
+	if len(settings) == 0 {
+		return ""
+	}
+	return ipcauth.UpCommand(strings.Join(upFlags(settings), " "))
+}
+
+// guardedSummary names the change for the log.
+func guardedSummary(p GuardedSettings) string {
+	return fmt.Sprintf("%v for profile %q", oneShotArgs(guardedSettings(p)), p.ProfileName)
+}

client/ui/services/guarded_test.go

@@ -0,0 +1,349 @@
+//go:build !android && !ios && !freebsd && !js
+
+package services
+
+import (
+	"context"
+	"errors"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"google.golang.org/genproto/googleapis/rpc/errdetails"
+	"google.golang.org/grpc"
+	"google.golang.org/grpc/codes"
+	gstatus "google.golang.org/grpc/status"
+
+	"github.com/netbirdio/netbird/client/internal/elevate"
+	"github.com/netbirdio/netbird/client/internal/ipcauth"
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+// A Unix socket, so the daemon address is one that carries a caller's identity and
+// elevation is worth offering at all: see Settings.canElevate.
+const testDaemonAddr = "unix:///var/run/netbird.sock"
+
+// storedManagementURL is what the stub daemon already holds, so that a request
+// naming a different one is a change: see Settings.guardedChanges.
+const storedManagementURL = "https://stored.example.com"
+
+// stubElevator stands in for the platform's prompt: it records what would have run
+// and answers with a fixed outcome.
+type stubElevator struct {
+	outcome   error
+	available bool
+	calls     [][]string
+}
+
+func (e *stubElevator) Run(_ context.Context, args ...string) error {
+	e.calls = append(e.calls, args)
+	return e.outcome
+}
+
+func (e *stubElevator) Available() bool { return e.available }
+
+// stubDaemon implements only the RPCs under test. The embedded interface is nil, so
+// any other call panics rather than passing quietly.
+type stubDaemon struct {
+	proto.DaemonServiceClient
+	setConfig func(*proto.SetConfigRequest) error
+	// stored is what GetConfig reports, which is what a refused request's guarded
+	// settings are compared against.
+	stored   *proto.GetConfigResponse
+	requests []*proto.SetConfigRequest
+}
+
+func (d *stubDaemon) SetConfig(_ context.Context, in *proto.SetConfigRequest, _ ...grpc.CallOption) (*proto.SetConfigResponse, error) {
+	d.requests = append(d.requests, in)
+	if err := d.setConfig(in); err != nil {
+		return nil, err
+	}
+	return &proto.SetConfigResponse{}, nil
+}
+
+func (d *stubDaemon) GetConfig(_ context.Context, _ *proto.GetConfigRequest, _ ...grpc.CallOption) (*proto.GetConfigResponse, error) {
+	return d.stored, nil
+}
+
+type stubConn struct{ client proto.DaemonServiceClient }
+
+func (c stubConn) Client() (proto.DaemonServiceClient, error) { return c.client, nil }
+
+// privilegeRefusal is the error the daemon raises for a change it restricts to
+// root, detail and all: see server.privilegeError.
+func privilegeRefusal(t *testing.T) error {
+	t.Helper()
+
+	st, err := gstatus.New(codes.PermissionDenied, "Changing the management URL requires root.").
+		WithDetails(&errdetails.ErrorInfo{
+			Reason: ipcauth.ErrorReasonPrivilegeRequired,
+			Domain: ipcauth.ErrorDomain,
+			Metadata: map[string]string{
+				ipcauth.ErrorMetaSummary: "Changing the management URL requires root.",
+				ipcauth.ErrorMetaCommand: "sudo netbird down; sudo netbird up -m https://mgmt.example.com",
+			},
+		})
+	require.NoError(t, err, "build the refusal detail")
+	return st.Err()
+}
+
+func settingsWithElevation(t *testing.T, outcome error) (*Settings, *stubElevator) {
+	t.Helper()
+
+	elev := &stubElevator{outcome: outcome, available: true}
+	return &Settings{daemonAddr: testDaemonAddr, elevator: elev}, elev
+}
+
+// settingsRefusingOnce returns a Settings whose daemon refuses the first SetConfig
+// for want of privileges and accepts anything after it. Its stored config holds
+// another management server and no SSH grants, so a request naming either is a
+// change rather than a restatement.
+func settingsRefusingOnce(t *testing.T, elev *stubElevator) (*Settings, *stubDaemon) {
+	t.Helper()
+
+	refusal := privilegeRefusal(t)
+	daemon := &stubDaemon{stored: &proto.GetConfigResponse{ManagementUrl: storedManagementURL}}
+	daemon.setConfig = func(*proto.SetConfigRequest) error {
+		if len(daemon.requests) == 1 {
+			return refusal
+		}
+		return nil
+	}
+	return &Settings{conn: stubConn{client: daemon}, daemonAddr: testDaemonAddr, elevator: elev}, daemon
+}
+
+func TestSetGuardedSettingsPassesOnlyTheChangedSettings(t *testing.T) {
+	s, elev := settingsWithElevation(t, nil)
+
+	root := true
+	outcome, err := s.SetGuardedSettings(context.Background(), GuardedSettings{
+		ProfileName:   "work",
+		Username:      "vma",
+		EnableSSHRoot: &root,
+	})
+	require.NoError(t, err)
+	assert.False(t, outcome.Declined, "the prompt was answered")
+
+	want := []string{
+		"--" + FlagApplyPrivilegedSettings,
+		"--" + FlagDaemonAddr, testDaemonAddr,
+		"--" + FlagProfile, "work",
+		"--" + FlagUser, "vma",
+		"--" + FlagEnableSSHRoot + "=true",
+	}
+	require.Len(t, elev.calls, 1, "one prompt for one change")
+	assert.Equal(t, want, elev.calls[0], "elevated arguments")
+}
+
+// Turning a setting off has to be as explicit as turning it on: a bare flag would
+// read as "on" to the one-shot's parser.
+func TestSetGuardedSettingsSpellsOutFalse(t *testing.T) {
+	s, elev := settingsWithElevation(t, nil)
+
+	off := false
+	_, err := s.SetGuardedSettings(context.Background(), GuardedSettings{
+		ProfileName:      "default",
+		ServerSSHAllowed: &off,
+		DisableSSHAuth:   &off,
+	})
+	require.NoError(t, err)
+
+	args := elev.calls[0]
+	assert.Contains(t, args, "--"+FlagAllowServerSSH+"=false", "the setting being switched off")
+	assert.Contains(t, args, "--"+FlagDisableSSHAuth+"=false", "the setting being switched off")
+	assert.NotContains(t, args, "--"+FlagEnableSSHRoot+"=false", "no flag for a setting nobody touched")
+}
+
+func TestSetGuardedSettingsPassesTheManagementURL(t *testing.T) {
+	s, elev := settingsWithElevation(t, nil)
+
+	_, err := s.SetGuardedSettings(context.Background(), GuardedSettings{
+		ProfileName:   "default",
+		ManagementURL: "https://mgmt.example.com:33073",
+	})
+	require.NoError(t, err)
+
+	assert.Contains(t, elev.calls[0], "--"+FlagManagementURL+"=https://mgmt.example.com:33073",
+		"the management URL to point the profile at")
+}
+
+func TestSetGuardedSettingsWithoutASettingDoesNotElevate(t *testing.T) {
+	s, elev := settingsWithElevation(t, nil)
+
+	_, err := s.SetGuardedSettings(context.Background(), GuardedSettings{ProfileName: "default"})
+
+	require.Error(t, err, "nothing to apply is not something to prompt for")
+	assert.Empty(t, elev.calls, "no prompt at all")
+}
+
+// A declined prompt is the one ending that is not an error: reporting it as one
+// would have every cancelled prompt logged as a failure.
+func TestSetGuardedSettingsReportsADeclinedPromptAsAnOutcome(t *testing.T) {
+	s, _ := settingsWithElevation(t, elevate.ErrDeclined)
+
+	root := true
+	outcome, err := s.SetGuardedSettings(context.Background(), GuardedSettings{
+		ProfileName:   "default",
+		EnableSSHRoot: &root,
+	})
+
+	require.NoError(t, err, "the user was asked and answered; nothing went wrong")
+	assert.True(t, outcome.Declined, "nothing was applied")
+}
+
+func TestSetGuardedSettingsMapsFailures(t *testing.T) {
+	tests := []struct {
+		name     string
+		outcome  error
+		wantCode string
+	}{
+		{
+			// Nothing to raise a prompt with: the user needs the command.
+			name:     "no mechanism falls back to the command",
+			outcome:  elevate.ErrUnavailable,
+			wantCode: CodeElevationUnavailable,
+		},
+		{
+			name:     "a failed run falls back to the command",
+			outcome:  errors.New("elevated netbird exited with 1"),
+			wantCode: CodeElevationFailed,
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			s, _ := settingsWithElevation(t, tt.outcome)
+
+			root := true
+			_, err := s.SetGuardedSettings(context.Background(), GuardedSettings{
+				ProfileName:   "default",
+				EnableSSHRoot: &root,
+			})
+
+			var clientErr *ClientError
+			require.ErrorAs(t, err, &clientErr, "the frontend needs a code to act on")
+			assert.Equal(t, tt.wantCode, clientErr.Code, "error code")
+			assert.Contains(t, clientErr.Command, "--"+FlagEnableSSHRoot+"=true",
+				"the setting in the fallback command")
+			assert.Contains(t, clientErr.Command, "netbird up", "the fallback command")
+		})
+	}
+}
+
+// Changing the management URL is only privileged while the host runs the SSH
+// server, which no control can know up front, so the refusal is what triggers the
+// prompt. The original request goes again afterwards, so the fields the one-shot
+// does not understand are applied too.
+func TestSetConfigElevatesAfterARefusalAndRetries(t *testing.T) {
+	elev := &stubElevator{available: true}
+	s, daemon := settingsRefusingOnce(t, elev)
+
+	mtu := int64(1280)
+	outcome, err := s.SetConfig(context.Background(), SetConfigParams{
+		ProfileName:   "default",
+		ManagementURL: "https://mgmt.example.com",
+		MTU:           &mtu,
+	})
+	require.NoError(t, err)
+	assert.False(t, outcome.Declined, "the prompt was answered")
+
+	require.Len(t, elev.calls, 1, "one prompt")
+	assert.Contains(t, elev.calls[0], "--"+FlagManagementURL+"=https://mgmt.example.com",
+		"the guarded part of the request")
+	require.Len(t, daemon.requests, 2, "the refused request and the retry")
+	assert.Equal(t, mtu, daemon.requests[1].GetMtu(),
+		"the retry carries the rest of the request, which the one-shot does not understand")
+}
+
+func TestSetConfigDoesNotRetryWhenTheUserDeclines(t *testing.T) {
+	elev := &stubElevator{outcome: elevate.ErrDeclined, available: true}
+	s, daemon := settingsRefusingOnce(t, elev)
+
+	outcome, err := s.SetConfig(context.Background(), SetConfigParams{
+		ProfileName:   "default",
+		ManagementURL: "https://mgmt.example.com",
+	})
+
+	require.NoError(t, err, "a declined prompt is not an error")
+	assert.True(t, outcome.Declined, "nothing was applied")
+	assert.Len(t, daemon.requests, 1, "only the refused request")
+}
+
+// With no prompt to raise, the refusal is reported as the daemon wrote it, which is
+// the guidance that was there before elevation existed.
+func TestSetConfigReportsTheRefusalWhenItCannotElevate(t *testing.T) {
+	elev := &stubElevator{available: false}
+	s, _ := settingsRefusingOnce(t, elev)
+
+	_, err := s.SetConfig(context.Background(), SetConfigParams{
+		ProfileName:   "default",
+		ManagementURL: "https://mgmt.example.com",
+	})
+
+	var clientErr *ClientError
+	require.ErrorAs(t, err, &clientErr)
+	assert.Equal(t, "privilege_required", clientErr.Code, "error code")
+	assert.Contains(t, clientErr.Command, "netbird up -m https://mgmt.example.com",
+		"the daemon's own command")
+	assert.Empty(t, elev.calls, "no prompt where there is none to raise")
+}
+
+// One authorization must buy only the change the user made. A settings form
+// submits every field it holds, so most of a refused request restates what the
+// daemon already has, and elevating those too would apply a guarded setting the
+// user never touched — a value gone stale since the form loaded above all.
+func TestSetConfigElevatesOnlyTheGuardedSettingsThatChange(t *testing.T) {
+	elev := &stubElevator{available: true}
+	s, _ := settingsRefusingOnce(t, elev)
+
+	on, off := true, false
+	_, err := s.SetConfig(context.Background(), SetConfigParams{
+		ProfileName:      "default",
+		ManagementURL:    storedManagementURL,
+		ServerSSHAllowed: &off,
+		EnableSSHRoot:    &off,
+		DisableSSHAuth:   &on,
+	})
+	require.NoError(t, err)
+
+	require.Len(t, elev.calls, 1, "one prompt")
+	args := elev.calls[0]
+	assert.Contains(t, args, "--"+FlagDisableSSHAuth+"=true", "the setting that changes")
+	assert.NotContains(t, args, "--"+FlagManagementURL+"="+storedManagementURL,
+		"a management URL the daemon already holds")
+	assert.NotContains(t, args, "--"+FlagAllowServerSSH+"=false", "a setting already off")
+	assert.NotContains(t, args, "--"+FlagEnableSSHRoot+"=false", "a setting already off")
+}
+
+// A request that changes no guarded setting has nothing an elevated run could
+// apply, so the refusal must have come from somewhere a prompt cannot reach.
+func TestSetConfigDoesNotElevateWhenNoGuardedSettingChanges(t *testing.T) {
+	elev := &stubElevator{available: true}
+	s, _ := settingsRefusingOnce(t, elev)
+
+	off := false
+	_, err := s.SetConfig(context.Background(), SetConfigParams{
+		ProfileName:      "default",
+		ManagementURL:    storedManagementURL,
+		ServerSSHAllowed: &off,
+	})
+
+	var clientErr *ClientError
+	require.ErrorAs(t, err, &clientErr)
+	assert.Equal(t, "privilege_required", clientErr.Code, "error code")
+	assert.Empty(t, elev.calls, "no prompt for a change nobody made")
+}
+
+// A refusal with nothing in the request the one-shot could apply: the daemon
+// cannot see who is calling, and being root would not help either.
+func TestSetConfigReportsARefusalWithNothingToElevate(t *testing.T) {
+	elev := &stubElevator{available: true}
+	s, _ := settingsRefusingOnce(t, elev)
+
+	_, err := s.SetConfig(context.Background(), SetConfigParams{ProfileName: "default"})
+
+	var clientErr *ClientError
+	require.ErrorAs(t, err, &clientErr)
+	assert.Equal(t, "privilege_required", clientErr.Code, "error code")
+	assert.Empty(t, elev.calls, "no prompt")
+}

client/ui/services/oneshot.go

@@ -0,0 +1,239 @@
+//go:build !android && !ios && !freebsd && !js
+
+package services
+
+import (
+	"context"
+	"errors"
+	"flag"
+	"fmt"
+	"os"
+	"strconv"
+	"time"
+
+	gstatus "google.golang.org/grpc/status"
+
+	"github.com/netbirdio/netbird/client/internal/elevate"
+	"github.com/netbirdio/netbird/client/internal/profilemanager"
+	"github.com/netbirdio/netbird/client/proto"
+	"github.com/netbirdio/netbird/util"
+)
+
+// The other end of SetGuardedSettings: the mode this binary runs itself in,
+// elevated, to apply the settings the daemon restricts to root/administrator.
+//
+// Both ends are here on purpose. What may be changed this way is an allowlist, and
+// an allowlist declared twice is one that will eventually disagree with itself, so
+// the arguments are rendered and parsed from a single table: guardedFields. Adding
+// a setting is one row; nothing generic passes through, and no field outside the
+// table can be reached with an elevated request no matter what lands on the command
+// line.
+
+// oneShotTimeout bounds the whole one-shot: connect, one RPC, exit. Generous
+// because the user has just waited for an authentication dialog, and a failure here
+// costs them the entire round trip.
+const oneShotTimeout = 30 * time.Second
+
+// Exit codes the parent reads where the platform gives it one.
+const (
+	exitOK      = 0
+	exitFailure = 1
+	exitUsage   = 2
+)
+
+// guardedField is one setting the one-shot understands, in the two spellings it
+// needs and with the two halves of its plumbing.
+type guardedField struct {
+	// flag names it on the one-shot's command line.
+	flag  string
+	usage string
+	// read returns the value to send and whether the caller asked for this setting
+	// at all.
+	read func(GuardedSettings) (string, bool)
+	// write parses a value from the command line onto the request. It is the only
+	// thing that validates the value, so it fails on anything it does not
+	// recognise rather than guessing.
+	write func(*proto.SetConfigRequest, string) error
+	// up renders the equivalent `netbird up` flag, for the fallback command shown
+	// when there is no prompt to raise.
+	up func(value string) string
+}
+
+var guardedFields = []guardedField{
+	{
+		flag:  FlagManagementURL,
+		usage: "Management server the profile registers with.",
+		read:  func(p GuardedSettings) (string, bool) { return p.ManagementURL, p.ManagementURL != "" },
+		write: func(req *proto.SetConfigRequest, value string) error {
+			// Parsed with the config layer's own parser, so what the elevated run
+			// accepts cannot drift from what the daemon would store.
+			if _, err := profilemanager.ParseServiceURL("Management URL", value); err != nil {
+				return err
+			}
+			req.ManagementUrl = value
+			return nil
+		},
+		// The daemon names this one as `-m <url>` in its own refusals.
+		up: func(value string) string { return "-m " + value },
+	},
+	boolField(FlagAllowServerSSH, "Run the NetBird SSH server.",
+		func(p GuardedSettings) *bool { return p.ServerSSHAllowed },
+		func(req *proto.SetConfigRequest, v *bool) { req.ServerSSHAllowed = v }),
+	boolField(FlagEnableSSHRoot, "Allow SSH sessions to privileged accounts.",
+		func(p GuardedSettings) *bool { return p.EnableSSHRoot },
+		func(req *proto.SetConfigRequest, v *bool) { req.EnableSSHRoot = v }),
+	boolField(FlagDisableSSHAuth, "Accept SSH sessions without authentication.",
+		func(p GuardedSettings) *bool { return p.DisableSSHAuth },
+		func(req *proto.SetConfigRequest, v *bool) { req.DisableSSHAuth = v }),
+}
+
+// fieldValue is a flag that remembers whether it was given, and requires a value:
+// the renderer always writes one, so a bare flag is a caller that got it wrong.
+type fieldValue struct {
+	set   bool
+	value string
+}
+
+func (v *fieldValue) String() string {
+	if v == nil {
+		return ""
+	}
+	return v.value
+}
+
+func (v *fieldValue) Set(value string) error {
+	v.set, v.value = true, value
+	return nil
+}
+
+// boolField describes a setting that is on or off. The value is always spelled out,
+// so that turning a setting off is as unambiguous as turning it on and a flag with
+// no value is a mistake rather than an "on".
+func boolField(
+	name, usage string,
+	read func(GuardedSettings) *bool,
+	write func(*proto.SetConfigRequest, *bool),
+) guardedField {
+	return guardedField{
+		flag:  name,
+		usage: usage,
+		read: func(p GuardedSettings) (string, bool) {
+			value := read(p)
+			if value == nil {
+				return "", false
+			}
+			return strconv.FormatBool(*value), true
+		},
+		write: func(req *proto.SetConfigRequest, value string) error {
+			parsed, err := strconv.ParseBool(value)
+			if err != nil {
+				return fmt.Errorf("parse %q as a boolean: %w", value, err)
+			}
+			write(req, &parsed)
+			return nil
+		},
+		up: func(value string) string { return "--" + name + "=" + value },
+	}
+}
+
+// IsPrivilegedSettingsRun reports whether this process was started as the one-shot.
+// The flag is a marker rather than a value, so only the bare forms count: reading a
+// value would mean "--flag=false" started it too.
+func IsPrivilegedSettingsRun(args []string) bool {
+	for _, arg := range args {
+		if arg == "--"+FlagApplyPrivilegedSettings || arg == "-"+FlagApplyPrivilegedSettings {
+			return true
+		}
+	}
+	return false
+}
+
+// RunPrivilegedSettings applies the requested settings and returns the process exit
+// code. connect dials the daemon, which is the caller's business because only it
+// knows how this build talks to it.
+//
+// Everything it reports goes to stderr, which is what the parent captures where the
+// platform lets it. On success it says so on standard output, because macOS gives
+// the parent no exit status to read: see elevate.AppliedMarker.
+func RunPrivilegedSettings(args []string, connect func(addr string) (proto.DaemonServiceClient, error)) int {
+	fs := flag.NewFlagSet("netbird-ui --"+FlagApplyPrivilegedSettings, flag.ContinueOnError)
+	fs.Bool(FlagApplyPrivilegedSettings, false, "Apply the settings the daemon restricts to root/administrator and exit.")
+	daemonAddr := fs.String(FlagDaemonAddr, "", "Daemon gRPC address: unix:///path, npipe://name or tcp://host:port")
+	logLevel := fs.String(FlagLogLevel, "info", "Log level: trace|debug|info|warn|error.")
+	profile := fs.String(FlagProfile, "", "Profile to change.")
+	username := fs.String(FlagUser, "", "Owner of the profile.")
+
+	values := make([]fieldValue, len(guardedFields))
+	for i, field := range guardedFields {
+		fs.Var(&values[i], field.flag, field.usage)
+	}
+
+	if err := fs.Parse(args); err != nil {
+		return exitUsage
+	}
+
+	if err := util.InitLog(*logLevel, "console"); err != nil {
+		fmt.Fprintf(os.Stderr, "init log: %v\n", err)
+		return exitFailure
+	}
+
+	req, err := privilegedRequest(*profile, *username, values)
+	if err != nil {
+		fmt.Fprintf(os.Stderr, "%v\n", err)
+		return exitUsage
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), oneShotTimeout)
+	defer cancel()
+
+	if err := applyPrivilegedSettings(ctx, *daemonAddr, req, connect); err != nil {
+		fmt.Fprintf(os.Stderr, "apply settings: %v\n", err)
+		return exitFailure
+	}
+
+	fmt.Fprintln(os.Stdout, elevate.AppliedMarker)
+	return exitOK
+}
+
+// privilegedRequest builds the request from the flags that were given, and refuses
+// one that asks for nothing.
+func privilegedRequest(profile, username string, values []fieldValue) (*proto.SetConfigRequest, error) {
+	req := &proto.SetConfigRequest{ProfileName: profile, Username: username}
+
+	given := 0
+	for i, field := range guardedFields {
+		if !values[i].set {
+			continue
+		}
+		if err := field.write(req, values[i].value); err != nil {
+			return nil, fmt.Errorf("--%s: %w", field.flag, err)
+		}
+		given++
+	}
+	if given == 0 {
+		return nil, errors.New("no setting to apply")
+	}
+	return req, nil
+}
+
+func applyPrivilegedSettings(
+	ctx context.Context,
+	daemonAddr string,
+	req *proto.SetConfigRequest,
+	connect func(addr string) (proto.DaemonServiceClient, error),
+) error {
+	client, err := connect(daemonAddr)
+	if err != nil {
+		return err
+	}
+	if _, err := client.SetConfig(ctx, req); err != nil {
+		// Unwrapped: the daemon's message is written for a person, and a refusal
+		// elevation cannot fix has to say so where the parent can read it off
+		// stderr.
+		return errors.New(gstatus.Convert(err).Message())
+	}
+	return nil
+}
+
+// interface guard: the one-shot's flags are flag.Value.
+var _ flag.Value = (*fieldValue)(nil)

client/ui/services/oneshot_test.go

@@ -0,0 +1,151 @@
+//go:build !android && !ios && !freebsd && !js
+
+package services
+
+import (
+	"flag"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/proto"
+)
+
+func TestIsPrivilegedSettingsRun(t *testing.T) {
+	tests := []struct {
+		name string
+		args []string
+		want bool
+	}{
+		{name: "no arguments"},
+		{name: "double dash", args: []string{"--" + FlagApplyPrivilegedSettings}, want: true},
+		{name: "single dash", args: []string{"-" + FlagApplyPrivilegedSettings}, want: true},
+		{
+			name: "among other flags",
+			args: []string{"--daemon-addr", "unix:///tmp/x.sock", "--" + FlagApplyPrivilegedSettings},
+			want: true,
+		},
+		// A marker, not a value: the caller never passes one, and reading a value
+		// would mean "--flag=false" started the one-shot too.
+		{name: "with a value", args: []string{"--" + FlagApplyPrivilegedSettings + "=true"}},
+		{name: "unrelated flags", args: []string{"--log-level", "debug"}},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			assert.Equal(t, tt.want, IsPrivilegedSettingsRun(tt.args), "args %v", tt.args)
+		})
+	}
+}
+
+// What SetGuardedSettings renders has to be what the one-shot reads back, for every
+// setting in the table. This is the property that keeps the two ends of an allowlist
+// from drifting, so it is checked field by field rather than by example.
+func TestGuardedFieldsRoundTrip(t *testing.T) {
+	on, off := true, false
+	tests := []struct {
+		name     string
+		settings GuardedSettings
+		want     func(*testing.T, *proto.SetConfigRequest)
+	}{
+		{
+			name:     "management url",
+			settings: GuardedSettings{ManagementURL: "https://mgmt.example.com:33073"},
+			want: func(t *testing.T, req *proto.SetConfigRequest) {
+				assert.Equal(t, "https://mgmt.example.com:33073", req.GetManagementUrl())
+			},
+		},
+		{
+			name:     "ssh server on",
+			settings: GuardedSettings{ServerSSHAllowed: &on},
+			want: func(t *testing.T, req *proto.SetConfigRequest) {
+				require.NotNil(t, req.ServerSSHAllowed)
+				assert.True(t, *req.ServerSSHAllowed)
+			},
+		},
+		{
+			name:     "ssh root off",
+			settings: GuardedSettings{EnableSSHRoot: &off},
+			want: func(t *testing.T, req *proto.SetConfigRequest) {
+				require.NotNil(t, req.EnableSSHRoot, "an explicit false must survive, not read as absent")
+				assert.False(t, *req.EnableSSHRoot)
+			},
+		},
+		{
+			name:     "ssh auth off",
+			settings: GuardedSettings{DisableSSHAuth: &on},
+			want: func(t *testing.T, req *proto.SetConfigRequest) {
+				require.NotNil(t, req.DisableSSHAuth)
+				assert.True(t, *req.DisableSSHAuth)
+			},
+		},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			req := parseRendered(t, tt.settings)
+			tt.want(t, req)
+		})
+	}
+}
+
+// A setting nobody asked about must not arrive at the daemon at all: sending its
+// zero value would change it.
+func TestGuardedFieldsCarryOnlyWhatWasAsked(t *testing.T) {
+	on := true
+	req := parseRendered(t, GuardedSettings{ProfileName: "work", EnableSSHRoot: &on})
+
+	assert.Equal(t, "work", req.GetProfileName(), "profile")
+	require.NotNil(t, req.EnableSSHRoot)
+	assert.Nil(t, req.ServerSSHAllowed, "untouched setting")
+	assert.Nil(t, req.DisableSSHAuth, "untouched setting")
+	assert.Empty(t, req.GetManagementUrl(), "untouched setting")
+}
+
+func TestPrivilegedRequestRejectsAnEmptyChange(t *testing.T) {
+	_, err := privilegedRequest("default", "vma", make([]fieldValue, len(guardedFields)))
+	require.Error(t, err, "nothing to apply is not a request worth sending as root")
+}
+
+// A value the table cannot parse is refused rather than guessed at.
+func TestPrivilegedRequestRejectsAnUnparseableValue(t *testing.T) {
+	values := make([]fieldValue, len(guardedFields))
+	for i, field := range guardedFields {
+		if field.flag != FlagEnableSSHRoot {
+			continue
+		}
+		require.NoError(t, values[i].Set("perhaps"))
+	}
+
+	_, err := privilegedRequest("default", "vma", values)
+	require.Error(t, err)
+	assert.Contains(t, err.Error(), FlagEnableSSHRoot, "which flag was wrong")
+}
+
+// parseRendered puts the settings through both ends: rendered as the arguments the
+// elevated process is given, then parsed by a flag set registered from the same
+// table, which is what the one-shot itself parses them with. Anything hand-rolled
+// here would pin down a parser nothing uses.
+func parseRendered(t *testing.T, p GuardedSettings) *proto.SetConfigRequest {
+	t.Helper()
+
+	rendered := guardedSettings(p)
+	require.NotEmpty(t, rendered, "nothing rendered for %+v", p)
+
+	args := make([]string, 0, len(rendered))
+	for _, setting := range rendered {
+		args = append(args, setting.arg)
+	}
+
+	fs := flag.NewFlagSet(t.Name(), flag.ContinueOnError)
+	values := make([]fieldValue, len(guardedFields))
+	for i, field := range guardedFields {
+		fs.Var(&values[i], field.flag, field.usage)
+	}
+	require.NoError(t, fs.Parse(args), "the one-shot's own flag set must accept %v", args)
+
+	req, err := privilegedRequest(p.ProfileName, p.Username, values)
+	require.NoError(t, err)
+	return req
+}

client/ui/services/profile.go

@@ -6,6 +6,8 @@ import (
 	"context"
 	"os/user"
 
+	log "github.com/sirupsen/logrus"
+
 	"github.com/netbirdio/netbird/client/internal/profilemanager"
 	"github.com/netbirdio/netbird/client/proto"
 )
@@ -151,11 +153,31 @@ func (s *Profiles) Remove(ctx context.Context, p ProfileRef) error {
 	if err != nil {
 		return err
 	}
-	_, err = cli.RemoveProfile(ctx, &proto.RemoveProfileRequest{
+	resp, err := cli.RemoveProfile(ctx, &proto.RemoveProfileRequest{
 		ProfileName: p.ProfileName,
 		Username:    p.Username,
 	})
-	return err
+	if err != nil {
+		return err
+	}
+
+	// The daemon deletes what it owns but runs as root, so it leaves the
+	// user-owned state file holding the account email behind (same split as
+	// Connection.Logout). Legacy profiles are keyed by name rather than by a
+	// generated ID, so a recreated profile of the same name would inherit the
+	// deleted one's email and offer it as the login_hint.
+	//
+	// Keyed on the ID the daemon resolved, not on the request handle: that may
+	// have been a display name or an ID prefix, which would name a different
+	// file (or none).
+	if id := resp.GetId(); id != "" {
+		if err := profilemanager.NewProfileManager().RemoveProfileState(id); err != nil {
+			// Non-fatal: the profile itself is gone.
+			log.Warnf("failed to remove profile state for %s: %v", id, err)
+		}
+	}
+
+	return nil
 }
 
 // Rename changes a profile's display name. The on-disk ID is unaffected, so

client/ui/services/settings.go

@@ -44,12 +44,19 @@ type Restrictions struct {
 }
 
 // Privilege tells the frontend whether this process may perform the changes the
-// daemon restricts to root/administrator, and carries the command for each so a
-// disabled control can show the way to do it.
+// daemon restricts to root/administrator, whether it can ask the operating
+// system for the privileges instead, and the command for each so a control that
+// can do neither can still show the way.
 type Privilege struct {
 	Privileged bool `json:"privileged"`
-	// Actor names what the operation requires ("root", "administrator privileges").
-	Actor string `json:"actor"`
+	// ActorKey identifies the principal the operation requires without wording it,
+	// so the frontend can name it in the user's language: see
+	// ipcauth.PrivilegedActorKey. The words are not sent, because English ones
+	// cannot be dropped into a translated sentence.
+	ActorKey string `json:"actorKey"`
+	// CanElevate reports whether a guarded control can offer to authorize the
+	// change through the platform's own prompt: see SetGuardedSettings.
+	CanElevate bool `json:"canElevate"`
 	// Commands equivalent to the settings the daemon guards, ready to copy.
 	AllowSSHServer string `json:"allowSshServer"`
 	EnableSSHRoot  string `json:"enableSshRoot"`
@@ -128,6 +135,9 @@ type Settings struct {
 	// daemonAddr is where the daemon listens, used to tell whether it runs as
 	// this user and would therefore authorize us: see Privilege.
 	daemonAddr string
+	// elevator raises the platform's privilege prompt when a change needs more
+	// rights than this process has.
+	elevator elevator
 }
 
 func NewSettings(conn DaemonConn, translator ErrorTranslator, prefs LanguagePreference, daemonAddr string) *Settings {
@@ -135,6 +145,7 @@ func NewSettings(conn DaemonConn, translator ErrorTranslator, prefs LanguagePref
 		conn:       conn,
 		classifier: errorClassifier{translator: translator, prefs: prefs},
 		daemonAddr: daemonAddr,
+		elevator:   osElevator{},
 	}
 }
 
@@ -180,10 +191,10 @@ func (s *Settings) GetConfig(ctx context.Context, p ConfigParams) (Config, error
 	}, nil
 }
 
-func (s *Settings) SetConfig(ctx context.Context, p SetConfigParams) error {
+func (s *Settings) SetConfig(ctx context.Context, p SetConfigParams) (SaveOutcome, error) {
 	cli, err := s.conn.Client()
 	if err != nil {
-		return err
+		return SaveOutcome{}, err
 	}
 	req := &proto.SetConfigRequest{
 		ProfileName:                   p.ProfileName,
@@ -215,19 +226,92 @@ func (s *Settings) SetConfig(ctx context.Context, p SetConfigParams) error {
 		SshJWTCacheTTL:                p.SSHJWTCacheTTL,
 	}
 	if _, err := cli.SetConfig(ctx, req); err != nil {
+		if _, refused := privilegeErrorInfo(err); refused {
+			return s.setConfigElevated(ctx, p, req, err)
+		}
 		// Classified so the frontend gets the daemon's guidance instead of the
-		// gRPC envelope, which is what a refused privileged change looks like.
-		return s.classifier.classify(err)
+		// gRPC envelope.
+		return SaveOutcome{}, s.classifier.classify(err)
 	}
-	return nil
+	return SaveOutcome{}, nil
+}
+
+// setConfigElevated answers a request the daemon refused for want of privileges by
+// asking the user to authorize it, and sending it again if they do. It is the same
+// offer the SSH settings make up front, for the changes a control cannot know are
+// guarded until it is told: repointing a profile at another management server is
+// only privileged while that host runs the SSH server.
+//
+// Two steps, because the elevated one-shot deliberately understands only the
+// settings the daemon guards: it applies those, and the original request then goes
+// through as this user, its privileged parts now asking for nothing that is not
+// already stored. Nothing was applied by the refused attempt — the daemon decides
+// before it writes — so there is no half-applied state to undo either way.
+func (s *Settings) setConfigElevated(ctx context.Context, p SetConfigParams, req *proto.SetConfigRequest, refusal error) (SaveOutcome, error) {
+	if !s.canElevate() {
+		return SaveOutcome{}, s.classifier.classify(refusal)
+	}
+
+	guarded, err := s.guardedChanges(ctx, p)
+	if err != nil {
+		log.Warnf("cannot tell which guarded settings this request changes: %v", err)
+		return SaveOutcome{}, s.classifier.classify(refusal)
+	}
+	if len(guardedSettings(guarded)) == 0 {
+		// Refused over something no prompt can settle, such as a control channel
+		// that carries no caller identity. Report the daemon's own guidance.
+		return SaveOutcome{}, s.classifier.classify(refusal)
+	}
+
+	outcome, err := s.SetGuardedSettings(ctx, guarded)
+	if err != nil || outcome.Declined {
+		return outcome, err
+	}
+
+	cli, err := s.conn.Client()
+	if err != nil {
+		return SaveOutcome{}, err
+	}
+	if _, err := cli.SetConfig(ctx, req); err != nil {
+		return SaveOutcome{}, s.classifier.classify(err)
+	}
+	return SaveOutcome{}, nil
+}
+
+// guardedChanges is the guarded part of a request, reduced to what it actually
+// changes.
+//
+// A settings form submits every field it holds, so a request restates values the
+// daemon already has. Carrying those into the elevated run would spend one
+// authorization on more than the user asked for, and a value that has gone stale
+// since the form was loaded would spend it on something they never asked about.
+func (s *Settings) guardedChanges(ctx context.Context, p SetConfigParams) (GuardedSettings, error) {
+	stored, err := s.GetConfig(ctx, ConfigParams{ProfileName: p.ProfileName, Username: p.Username})
+	if err != nil {
+		return GuardedSettings{}, fmt.Errorf("read the stored config: %w", err)
+	}
+
+	guarded := GuardedSettings{
+		ProfileName:      p.ProfileName,
+		Username:         p.Username,
+		ServerSSHAllowed: changedFlag(p.ServerSSHAllowed, stored.ServerSSHAllowed),
+		EnableSSHRoot:    changedFlag(p.EnableSSHRoot, stored.EnableSSHRoot),
+		DisableSSHAuth:   changedFlag(p.DisableSSHAuth, stored.DisableSSHAuth),
+	}
+	// An empty URL leaves the setting alone, which is the daemon's rule too.
+	if p.ManagementURL != "" && p.ManagementURL != stored.ManagementURL {
+		guarded.ManagementURL = p.ManagementURL
+	}
+	return guarded, nil
 }
 
 // Privilege reports whether this UI process could carry out the changes the
-// daemon restricts to root/administrator, and the command that performs the one
-// users hit in the SSH settings. It applies the daemon's own rule to what it can
-// see locally, so the frontend can present those controls as unavailable up front
-// instead of letting a save fail. No daemon round-trip, so it also works while the
-// daemon is down.
+// daemon restricts to root/administrator, whether it can instead ask the
+// operating system for the privileges when the user wants one of them, and the
+// command that performs the ones users hit in the SSH settings. It applies the
+// daemon's own rule to what it can see locally, so the frontend can decide up
+// front how to present those controls instead of letting a save fail. No daemon
+// round-trip, so it also works while the daemon is down.
 //
 // Being root or an elevated administrator is one way. The other is running as the
 // daemon's own user while the daemon is unprivileged, which the daemon accepts
@@ -237,26 +321,40 @@ func (s *Settings) SetConfig(ctx context.Context, p SetConfigParams) error {
 func (s *Settings) Privilege() Privilege {
 	id, err := ipcauth.CurrentProcessIdentity()
 	if err != nil {
-		// Fail closed: report unprivileged, which only ever disables controls.
+		// Fail closed: report unprivileged, which only ever asks for more.
 		log.Warnf("cannot read this process's identity, treating it as unprivileged: %v", err)
-		return newPrivilege(false)
+		return s.newPrivilege(false)
 	}
 	if id.IsPrivileged() {
-		return newPrivilege(true)
+		return s.newPrivilege(true)
 	}
-	return newPrivilege(daemonaddr.DaemonRunsAsSelf(s.daemonAddr))
+	return s.newPrivilege(daemonaddr.DaemonRunsAsSelf(s.daemonAddr))
 }
 
-func newPrivilege(privileged bool) Privilege {
+func (s *Settings) newPrivilege(privileged bool) Privilege {
 	return Privilege{
 		Privileged:     privileged,
-		Actor:          ipcauth.PrivilegedActor(),
+		ActorKey:       ipcauth.PrivilegedActorKey(),
+		CanElevate:     s.canElevate(),
 		AllowSSHServer: ipcauth.UpCommand("--allow-server-ssh"),
 		EnableSSHRoot:  ipcauth.UpCommand("--enable-ssh-root"),
 		DisableSSHAuth: ipcauth.UpCommand("--disable-ssh-auth"),
 	}
 }
 
+// canElevate reports whether offering the platform's elevation prompt would get
+// the user anywhere. It needs a mechanism to raise the prompt with and a control
+// channel that tells the daemon who is calling: on loopback TCP the daemon
+// refuses these changes to everybody, root included, so a prompt there would
+// only waste the user's password.
+func (s *Settings) canElevate() bool {
+	if !daemonaddr.CarriesIdentity(s.daemonAddr) {
+		log.Debugf("not offering elevation: the daemon address %s carries no caller identity", s.daemonAddr)
+		return false
+	}
+	return s.elevator.Available()
+}
+
 func (s *Settings) GetRestrictions(ctx context.Context) (Restrictions, error) {
 	cli, err := s.conn.Client()
 	if err != nil {
@@ -289,6 +387,15 @@ func (s *Settings) GetRestrictions(ctx context.Context) (Restrictions, error) {
 	return r, nil
 }
 
+// changedFlag returns requested only when it differs from what is stored, so a
+// setting the request merely restates is left out of the elevated run.
+func changedFlag(requested *bool, stored bool) *bool {
+	if requested == nil || *requested == stored {
+		return nil
+	}
+	return requested
+}
+
 func applyMDMRestrictions(mdm *MDMFields, cfgResp *proto.GetConfigResponse) {
 	managed := cfgResp.GetMDMManagedFields()
 	if len(managed) == 0 {

client/ui/tray.go

@@ -44,7 +44,7 @@ type TrayServices struct {
 	Profiles        *services.Profiles
 	Networks        *services.Networks
 	DaemonFeed      *services.DaemonFeed
-	Notifier        *notifications.NotificationService
+	Notifier        *Notifier
 	Update          *services.Update
 	ProfileSwitcher *services.ProfileSwitcher
 	WindowManager   *services.WindowManager

client/ui/tray_click_linux.go

@@ -4,17 +4,26 @@ package main
 
 // bindTrayClick wires the tray icon's left-click handler on Linux.
 //
-// Both Linux click paths converge on Wails' linuxSystemTray.Activate, which
-// fires the registered clickHandler:
-//   - Real SNI hosts (KDE Plasma, Waybar, GNOME Shell + AppIndicator) invoke
-//     org.kde.StatusNotifierItem.Activate over D-Bus on left-click.
-//   - The in-process StatusNotifierWatcher + XEmbed host used on minimal WMs
-//     (Fluxbox, i3, dwm, OpenBox) maps a Button1 press to that same Activate
-//     call itself (xembed_host_linux.go), so it routes through the same hook.
-// Registering OnClick here therefore covers both paths with one handler — no
-// changes to the watcher or XEmbed C code are needed. Left-click now opens the
-// main window; right-click still opens the menu via Wails' default
-// SecondaryActivate→OpenMenu handler (and the XEmbed GTK popup on minimal WMs).
+// Expected behaviour per tray host:
+//
+//	Host                           Left click              Right click
+//	KDE Plasma, Waybar             main window (Activate)  menu (host-rendered)
+//	GNOME Shell + AppIndicator     menu only               menu only
+//	Minimal WMs via XEmbed host    main window (Activate)  XEmbed GTK popup
+//
+// OnClick fires only on org.kde.StatusNotifierItem.Activate — a real left
+// click. KDE/Waybar send it over D-Bus; the in-process XEmbed host
+// (xembed_host_linux.go) maps a Button1 press to the same Activate call.
+//
+// GNOME Shell + AppIndicator never sends Activate: it renders the dbusmenu
+// on ANY click and only reports the menu opening via dbusmenu
+// Event("opened"). Upstream Wails treated that event as a click, so on GNOME
+// both buttons raised the main window on top of the menu, and on KDE/Waybar
+// a right click raised it over the freshly opened menu. The netbirdio/wails
+// fork (go.mod replace) drops that heuristic: a menu open never fires
+// OnClick. On GNOME the main window is reached via the "Open NetBird" menu
+// entry; left-click-opens-window is not achievable there anyway, since the
+// host always opens the menu itself.
 //
 // We do NOT register OnDoubleClick: Wails' Linux SNI backend never fires it
 // (unlike Windows). And we deliberately skip AttachWindow — it plus Wails3's

client/ui/tray_notify.go

@@ -44,7 +44,7 @@ func safeSendNotification(send sendFn, what string, opts notifications.Notificat
 // notifyIfDaemonOutdated probes the daemon once and fires an OS toast when it
 // is reachable but too old for this UI. A probe error means the daemon isn't
 // reachable (not outdated), so it is left to the normal connection flow.
-func notifyIfDaemonOutdated(compat *services.Compat, notifier *notifications.NotificationService, loc *Localizer) {
+func notifyIfDaemonOutdated(compat *services.Compat, notifier *Notifier, loc *Localizer) {
 	ready, err := compat.DaemonReady(context.Background())
 	if err != nil {
 		log.Debugf("daemon compatibility probe: %v", err)

client/ui/tray_session.go

@@ -27,11 +27,10 @@ const (
 	finalWarningCountdownSeconds = 120
 )
 
-// handleSessionExpired notifies and brings the window forward so the frontend's /login route drives renewal.
+// handleSessionExpired notifies and brings the window forward so the user can reconnect.
 func (t *Tray) handleSessionExpired() {
 	t.notify(t.loc.T("notify.sessionExpired.title"), t.loc.T("notify.sessionExpired.body"), notifyIDSessionExpired)
 	if t.window != nil {
-		t.window.SetURL("/#/login")
 		t.window.Show()
 		t.window.Focus()
 	}
@@ -308,11 +307,7 @@ func (t *Tray) openSessionExtendFlow() {
 	}
 	seconds := int(time.Until(deadline).Seconds())
 	if seconds <= 0 {
-		if t.window != nil {
-			t.window.SetURL("/#/login")
-			t.window.Show()
-			t.window.Focus()
-		}
+		t.app.Event.Emit(services.EventTriggerLogin)
 		return
 	}
 	if t.svc.WindowManager == nil {

client/ui/tray_update.go

@@ -21,7 +21,7 @@ type trayUpdater struct {
 	app          *application.App
 	window       *application.WebviewWindow
 	update       *services.Update
-	notifier     *notifications.NotificationService
+	notifier     *Notifier
 	loc          *Localizer
 	onIconChange func()
 	// onMenuChange drives a full tray relayout: the update row lives in the
@@ -36,7 +36,7 @@ type trayUpdater struct {
 	progressWindowOpen bool
 }
 
-func newTrayUpdater(app *application.App, window *application.WebviewWindow, update *services.Update, notifier *notifications.NotificationService, loc *Localizer, onIconChange func(), onMenuChange func()) *trayUpdater {
+func newTrayUpdater(app *application.App, window *application.WebviewWindow, update *services.Update, notifier *Notifier, loc *Localizer, onIconChange func(), onMenuChange func()) *trayUpdater {
 	u := &trayUpdater{
 		app:          app,
 		window:       window,

e2e/agentnetwork/management_test.go

@@ -160,8 +160,19 @@ func TestSettingsRoundTrip(t *testing.T) {
 	assert.Equal(t, before.Cluster, flipped.Cluster, "cluster must be immutable across updates")
 	assert.Equal(t, before.Subdomain, flipped.Subdomain, "subdomain must be immutable across updates")
 
+	// A cluster different from the pinned one must be rejected; echoing the
+	// pinned one back is valid.
+	_, err = srv.UpdateSettings(ctx, api.AgentNetworkSettingsRequest{
+		Cluster:                ptr("attacker.cluster.invalid"),
+		EnableLogCollection:    before.EnableLogCollection,
+		EnablePromptCollection: before.EnablePromptCollection,
+		RedactPii:              before.RedactPii,
+	})
+	requireClientError(t, err)
+
 	// Restore the original toggles.
 	_, err = srv.UpdateSettings(ctx, api.AgentNetworkSettingsRequest{
+		Cluster:                ptr(before.Cluster),
 		EnableLogCollection:    before.EnableLogCollection,
 		EnablePromptCollection: before.EnablePromptCollection,
 		RedactPii:              before.RedactPii,

e2e/agentnetwork/settings_bootstrap_test.go

@@ -0,0 +1,114 @@
+//go:build e2e
+
+package agentnetwork
+
+import (
+	"context"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/e2e/harness"
+	"github.com/netbirdio/netbird/shared/management/http/api"
+)
+
+// harnessStartFresh boots a dedicated combined server with its own fresh
+// account and registers its teardown on t.
+func harnessStartFresh(ctx context.Context, t *testing.T) (*harness.Combined, error) {
+	t.Helper()
+	fresh, err := harness.StartCombined(ctx)
+	if err != nil {
+		return nil, err
+	}
+	t.Cleanup(func() { _ = fresh.Terminate(context.Background()) })
+	if _, err := fresh.Bootstrap(ctx); err != nil {
+		return nil, err
+	}
+	return fresh, nil
+}
+
+// TestSettingsBootstrapViaPut covers the settings-first bootstrap path on an
+// account that has never been bootstrapped: the GET reads as the defaults
+// with an empty cluster/subdomain/endpoint, a PUT without a cluster has
+// nothing to pin and fails, and a PUT carrying a cluster creates the row and
+// pins it immutably. The shared srv cannot provide that starting state (any
+// provider-creating test bootstraps it, and test order is deliberately not
+// relied on), so this boots a dedicated combined server — the image is
+// already built and cached by TestMain's StartCombined, so the extra cost is
+// one container start.
+func TestSettingsBootstrapViaPut(t *testing.T) {
+	ctx := context.Background()
+
+	fresh, err := harnessStartFresh(ctx, t)
+	require.NoError(t, err, "start dedicated combined server")
+
+	// Before agent-network bootstrap the settings read as the defaults, not
+	// as an error and not as a null body.
+	before, err := fresh.GetSettings(ctx)
+	require.NoError(t, err, "get settings on a fresh account must succeed")
+	assert.Empty(t, before.Cluster, "cluster must be empty before bootstrap")
+	assert.Empty(t, before.Subdomain, "subdomain must be empty before bootstrap")
+	assert.Empty(t, before.Endpoint, "endpoint must be empty before bootstrap, not a bare dot")
+	assert.True(t, before.EnableLogCollection, "defaults must show log collection on, matching bootstrap")
+	assert.False(t, before.EnablePromptCollection, "defaults must show prompt collection off")
+
+	// A PUT without a cluster has nothing to pin the account to.
+	_, err = fresh.UpdateSettings(ctx, api.AgentNetworkSettingsRequest{
+		EnableLogCollection: true,
+	})
+	requireClientError(t, err)
+
+	// A PUT carrying a cluster bootstraps the account and applies the
+	// mutable fields from the same request. Every toggle is set away from
+	// its bootstrap default so each assertion can actually fail.
+	const cluster = "e2e.bootstrap.netbird.selfhosted"
+	bootstrapped, err := fresh.UpdateSettings(ctx, api.AgentNetworkSettingsRequest{
+		Cluster:                ptr(cluster),
+		EnableLogCollection:    false,
+		EnablePromptCollection: true,
+		RedactPii:              true,
+	})
+	require.NoError(t, err, "bootstrap settings via PUT must succeed")
+	assert.Equal(t, cluster, bootstrapped.Cluster, "cluster must be pinned from the request")
+	require.NotEmpty(t, bootstrapped.Subdomain, "subdomain must be assigned at bootstrap")
+	assert.Equal(t, bootstrapped.Subdomain+"."+cluster, bootstrapped.Endpoint, "endpoint must combine subdomain and cluster")
+	assert.False(t, bootstrapped.EnableLogCollection, "log collection from the bootstrap request must override the default")
+	assert.True(t, bootstrapped.EnablePromptCollection, "prompt collection from the bootstrap request must apply")
+	assert.True(t, bootstrapped.RedactPii, "redact toggle from the bootstrap request must apply")
+
+	// The row is persisted: an independent read agrees on every field.
+	after, err := fresh.GetSettings(ctx)
+	require.NoError(t, err, "get settings after bootstrap must succeed")
+	assert.Equal(t, bootstrapped.Endpoint, after.Endpoint, "bootstrap must persist across reads")
+	assert.Equal(t, bootstrapped.EnableLogCollection, after.EnableLogCollection, "log collection must persist")
+	assert.Equal(t, bootstrapped.EnablePromptCollection, after.EnablePromptCollection, "prompt collection must persist")
+	assert.Equal(t, bootstrapped.RedactPii, after.RedactPii, "redact toggle must persist")
+
+	// Once bootstrapped, later updates may omit the cluster entirely.
+	persisted, err := fresh.UpdateSettings(ctx, api.AgentNetworkSettingsRequest{
+		EnableLogCollection:    true,
+		EnablePromptCollection: false,
+		RedactPii:              true,
+	})
+	require.NoError(t, err, "post-bootstrap update without cluster must succeed")
+	assert.Equal(t, cluster, persisted.Cluster, "omitted cluster must keep the pinned value")
+	assert.True(t, persisted.EnableLogCollection, "post-bootstrap toggle must apply")
+	assert.False(t, persisted.EnablePromptCollection, "post-bootstrap toggle must apply")
+
+	// The cluster is immutable: a different value is rejected rather than
+	// silently ignored, and the rejected update must not disturb anything.
+	_, err = fresh.UpdateSettings(ctx, api.AgentNetworkSettingsRequest{
+		Cluster:             ptr("other.cluster.invalid"),
+		EnableLogCollection: false,
+	})
+	requireClientError(t, err)
+
+	final, err := fresh.GetSettings(ctx)
+	require.NoError(t, err, "get settings after the rejected cluster change must succeed")
+	assert.Equal(t, persisted.Cluster, final.Cluster, "rejected update must not change the cluster")
+	assert.Equal(t, persisted.Endpoint, final.Endpoint, "rejected update must not change the endpoint")
+	assert.Equal(t, persisted.EnableLogCollection, final.EnableLogCollection, "rejected update must not apply its toggles")
+	assert.Equal(t, persisted.EnablePromptCollection, final.EnablePromptCollection, "rejected update must not apply its toggles")
+	assert.Equal(t, persisted.RedactPii, final.RedactPii, "rejected update must not apply its toggles")
+}

go.mod

@@ -114,7 +114,7 @@ require (
 	github.com/ti-mo/conntrack v0.5.1
 	github.com/ti-mo/netfilter v0.5.2
 	github.com/vmihailenco/msgpack/v5 v5.4.1
-	github.com/wailsapp/wails/v3 v3.0.0-alpha2.117
+	github.com/wailsapp/wails/v3 v3.0.0-beta.3
 	github.com/yusufpapurcu/wmi v1.2.4
 	github.com/zcalusic/sysinfo v1.1.3
 	go.opentelemetry.io/contrib/instrumentation/google.golang.org/grpc/otelgrpc v0.67.0
@@ -339,3 +339,5 @@ replace github.com/dexidp/dex => github.com/netbirdio/dex v0.244.1-0.20260716205
 replace github.com/dexidp/dex/api/v2 => github.com/netbirdio/dex/api/v2 v2.0.0-20260512110716-8d70ad8647c1
 
 replace github.com/mailru/easyjson => github.com/netbirdio/easyjson v0.9.0
+
+replace github.com/wailsapp/wails/v3 => github.com/netbirdio/wails/v3 v3.0.0-beta.3.0.20260803205919-ad21e92381f4

go.sum

@@ -490,6 +490,8 @@ github.com/netbirdio/service v0.0.0-20240911161631-f62744f42502 h1:3tHlFmhTdX9ax
 github.com/netbirdio/service v0.0.0-20240911161631-f62744f42502/go.mod h1:CIMRFEJVL+0DS1a3Nx06NaMn4Dz63Ng6O7dl0qH0zVM=
 github.com/netbirdio/signal-dispatcher/dispatcher v0.0.0-20250805121659-6b4ac470ca45 h1:ujgviVYmx243Ksy7NdSwrdGPSRNE3pb8kEDSpH0QuAQ=
 github.com/netbirdio/signal-dispatcher/dispatcher v0.0.0-20250805121659-6b4ac470ca45/go.mod h1:5/sjFmLb8O96B5737VCqhHyGRzNFIaN/Bu7ZodXc3qQ=
+github.com/netbirdio/wails/v3 v3.0.0-beta.3.0.20260803205919-ad21e92381f4 h1:UKztc3QjWvzU5DZk+uYaOWN0x62NSe/pkxuPvzqZIy4=
+github.com/netbirdio/wails/v3 v3.0.0-beta.3.0.20260803205919-ad21e92381f4/go.mod h1:BzATbK71VFikMMMCo434wAi0QcaI03P+xeaWgDvQvjw=
 github.com/netbirdio/wireguard-go v0.0.0-20260628102922-2834bebf6c1a h1:3CWK+yTvRKOcC0Q8VCTGy4l60TEb27CQVS7LkMxwjmw=
 github.com/netbirdio/wireguard-go v0.0.0-20260628102922-2834bebf6c1a/go.mod h1:rpwXGsirqLqN2L0JDJQlwOboGHmptD5ZD6T2VmcqhTw=
 github.com/nxadm/tail v1.4.4/go.mod h1:kenIhsEOeOJmVchQTgglprH7qJGnHDVpk1VPCcaMI8A=
@@ -660,8 +662,6 @@ github.com/vmihailenco/msgpack/v5 v5.4.1 h1:cQriyiUvjTwOHg8QZaPihLWeRAAVoCpE00IU
 github.com/vmihailenco/msgpack/v5 v5.4.1/go.mod h1:GaZTsDaehaPpQVyxrf5mtQlH+pc21PIudVV/E3rRQok=
 github.com/vmihailenco/tagparser/v2 v2.0.0 h1:y09buUbR+b5aycVFQs/g70pqKVZNBmxwAhO7/IwNM9g=
 github.com/vmihailenco/tagparser/v2 v2.0.0/go.mod h1:Wri+At7QHww0WTrCBeu4J6bNtoV6mEfg5OIWRZA9qds=
-github.com/wailsapp/wails/v3 v3.0.0-alpha2.117 h1:udyjqPG3AIgkod5QDR/WblCkpV8R86BFPSrsWxSyt5Y=
-github.com/wailsapp/wails/v3 v3.0.0-alpha2.117/go.mod h1:74WH2FScMsgucZvHHvv7eOefDXCm/CjuIxqhhZgPhKg=
 github.com/wlynxg/anet v0.0.5 h1:J3VJGi1gvo0JwZ/P1/Yc/8p63SoW98B5dHkYDmpgvvU=
 github.com/wlynxg/anet v0.0.5/go.mod h1:eay5PRQr7fIVAMbTbchTnO9gG65Hg/uYGdc7mguHxoA=
 github.com/x448/float16 v0.8.4 h1:qLwI1I70+NjRFUR3zs1JPUCgaCXSh3SW62uAKT1mSBM=

management/internals/modules/agentnetwork/handlers/handlers_test.go

@@ -17,6 +17,7 @@ import (
 
 	"github.com/netbirdio/netbird/management/internals/modules/agentnetwork"
 	agentNetworkTypes "github.com/netbirdio/netbird/management/internals/modules/agentnetwork/types"
+	"github.com/netbirdio/netbird/management/server/account"
 	nbcontext "github.com/netbirdio/netbird/management/server/context"
 	"github.com/netbirdio/netbird/management/server/permissions"
 	"github.com/netbirdio/netbird/management/server/store"
@@ -61,10 +62,23 @@ func newAgentNetworkHandlerFixture(t *testing.T) *agentNetworkHandlerFixture {
 		Return(true, context.Background(), nil).
 		AnyTimes()
 
-	manager := agentnetwork.NewManager(st, perms, nil, nil)
+	// Swallow activity events so the mutation paths (create/update/delete)
+	// are exercisable through the HTTP layer.
+	accounts := account.NewMockManager(ctrl)
+	accounts.EXPECT().
+		StoreEvent(gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any(), gomock.Any()).
+		AnyTimes()
+	accounts.EXPECT().
+		UpdateAccountPeers(gomock.Any(), gomock.Any(), gomock.Any()).
+		AnyTimes()
+
+	manager := agentnetwork.NewManager(st, perms, accounts, nil)
 	h := &handler{manager: manager}
 
 	router := mux.NewRouter()
+	router.HandleFunc("/agent-network/providers", h.createProvider).Methods("POST")
+	router.HandleFunc("/agent-network/providers/{providerId}", h.getProvider).Methods("GET")
+	router.HandleFunc("/agent-network/providers/{providerId}", h.updateProvider).Methods("PUT")
 	h.addPolicyEndpoints(router)
 	h.addConsumptionEndpoints(router)
 	h.addBudgetRuleEndpoints(router)

management/internals/modules/agentnetwork/handlers/providers_handler_test.go

@@ -1,7 +1,9 @@
 package handlers
 
 import (
+	"encoding/json"
 	"math"
+	nethttp "net/http"
 	"testing"
 
 	"github.com/stretchr/testify/assert"
@@ -51,3 +53,50 @@ func TestValidate_ModelRates(t *testing.T) {
 		assert.Error(t, validate(base(m), true), "case %q must be rejected", name)
 	}
 }
+
+// TestProviderHandler_UpdateReplacesFullState pins the update contract shared
+// with the other PUT endpoints: the request replaces the provider's mutable
+// state, so optional fields absent from the JSON land as their zero values.
+// The two exceptions are server-side: the api_key (a secret — omitted means
+// "not rotated") and the session keypair, both preserved by the manager. The
+// identity headers stay on the wire as explicit empty strings so a cleared
+// value round-trips.
+func TestProviderHandler_UpdateReplacesFullState(t *testing.T) {
+	f := newAgentNetworkHandlerFixture(t)
+
+	create := `{
+        "provider_id": "openai_api",
+        "name": "openai",
+        "upstream_url": "https://api.openai.com",
+        "api_key": "sk-test",
+        "enabled": true,
+        "metadata_disabled": true,
+        "skip_tls_verification": true,
+        "extra_values": {"x-portkey-config": "pc-prod-3f2a"},
+        "identity_header_user_id": "x-bf-dim-netbird_user_id",
+        "models": [{"id": "gpt-4o", "input_per_1k": 0.0025, "output_per_1k": 0.01}]
+    }`
+	rec := f.do(t, nethttp.MethodPost, "/agent-network/providers", create)
+	require.Equal(t, nethttp.StatusOK, rec.Code, "create must succeed: %s", rec.Body.String())
+
+	var created api.AgentNetworkProvider
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &created))
+
+	// Minimal update: only the required fields, no api_key. Everything
+	// optional must land as its zero value.
+	update := `{"provider_id": "openai_api", "name": "openai-renamed", "upstream_url": "https://api.openai.com", "enabled": true}`
+	rec = f.do(t, nethttp.MethodPut, "/agent-network/providers/"+created.Id, update)
+	require.Equal(t, nethttp.StatusOK, rec.Code, "update without api_key must succeed (key is preserved): %s", rec.Body.String())
+
+	var updated api.AgentNetworkProvider
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &updated))
+	assert.Equal(t, "openai-renamed", updated.Name, "sent field must apply")
+	assert.True(t, updated.Enabled, "sent field must apply")
+	assert.False(t, updated.MetadataDisabled, "omitted metadata_disabled must land as false — PUT replaces the full state")
+	assert.False(t, updated.SkipTlsVerification, "omitted skip_tls_verification must land as false")
+	assert.Nil(t, updated.ExtraValues, "omitted extra_values must be cleared")
+	assert.Equal(t, "", updated.IdentityHeaderUserId, "omitted identity header must be cleared yet stay on the wire")
+	assert.Empty(t, updated.Models, "omitted models must be cleared")
+	assert.Contains(t, rec.Body.String(), `"identity_header_user_id":""`,
+		"cleared identity header must round-trip as an explicit empty string")
+}

management/internals/modules/agentnetwork/handlers/settings_handler.go

@@ -2,7 +2,6 @@ package handlers
 
 import (
 	"encoding/json"
-	"errors"
 	"net/http"
 
 	"github.com/gorilla/mux"
@@ -11,19 +10,20 @@ import (
 	nbcontext "github.com/netbirdio/netbird/management/server/context"
 	"github.com/netbirdio/netbird/shared/management/http/api"
 	"github.com/netbirdio/netbird/shared/management/http/util"
-	"github.com/netbirdio/netbird/shared/management/status"
 )
 
 // addSettingsEndpoints registers the Agent Network settings routes. The
-// settings row is bootstrapped server-side on first provider create; GET reads
-// it and PUT updates the mutable collection toggles (cluster/subdomain stay
-// immutable).
+// settings row is bootstrapped server-side on first provider create or on the
+// first PUT carrying a cluster; GET reads it and PUT applies a partial update
+// of the mutable collection toggles (cluster/subdomain stay immutable).
 func (h *handler) addSettingsEndpoints(router *mux.Router) {
 	router.HandleFunc("/agent-network/settings", h.getSettings).Methods("GET", "OPTIONS")
 	router.HandleFunc("/agent-network/settings", h.updateSettings).Methods("PUT", "OPTIONS")
 }
 
-// updateSettings applies the collection toggles to the account's settings row.
+// updateSettings replaces the mutable settings fields on the account's row.
+// A request carrying a cluster bootstraps the row when the account doesn't
+// have one yet.
 func (h *handler) updateSettings(w http.ResponseWriter, r *http.Request) {
 	userAuth, err := nbcontext.GetUserAuthFromContext(r.Context())
 	if err != nil {
@@ -48,11 +48,9 @@ func (h *handler) updateSettings(w http.ResponseWriter, r *http.Request) {
 	util.WriteJSONObject(r.Context(), w, updated.ToAPIResponse())
 }
 
-// getSettings returns the account's agent-network settings. The settings
-// row is bootstrapped on first provider create, so freshly-onboarded
-// accounts have nothing to read. Rather than 404-ing in that case (which
-// the dashboard would have to special-case), return a JSON null with 200
-// so consumers can branch on the body alone.
+// getSettings returns the account's agent-network settings. Accounts that
+// haven't been bootstrapped yet read as the defaults with an empty cluster,
+// subdomain and endpoint; the manager synthesises that view.
 func (h *handler) getSettings(w http.ResponseWriter, r *http.Request) {
 	userAuth, err := nbcontext.GetUserAuthFromContext(r.Context())
 	if err != nil {
@@ -62,11 +60,6 @@ func (h *handler) getSettings(w http.ResponseWriter, r *http.Request) {
 
 	settings, err := h.manager.GetSettings(r.Context(), userAuth.AccountId, userAuth.UserId)
 	if err != nil {
-		var sErr *status.Error
-		if errors.As(err, &sErr) && sErr.Type() == status.NotFound {
-			util.WriteJSONObject(r.Context(), w, nil)
-			return
-		}
 		util.WriteError(r.Context(), err, w)
 		return
 	}

management/internals/modules/agentnetwork/handlers/settings_handler_test.go

@@ -0,0 +1,137 @@
+package handlers
+
+import (
+	"encoding/json"
+	"net/http"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/shared/management/http/api"
+)
+
+// TestSettingsHandler_GetUnbootstrappedReturnsDefaults pins the settings-read
+// convention shared with the account and DNS settings endpoints: settings
+// always read as a JSON object. Before bootstrap that object carries the
+// defaults with an empty cluster/subdomain/endpoint (the "not bootstrapped"
+// signal) and no timestamps — never a 404 and never the legacy null body.
+func TestSettingsHandler_GetUnbootstrappedReturnsDefaults(t *testing.T) {
+	f := newAgentNetworkHandlerFixture(t)
+
+	rec := f.do(t, http.MethodGet, "/agent-network/settings", "")
+	require.Equal(t, http.StatusOK, rec.Code,
+		"unbootstrapped account must read as 200 with defaults: got %d body=%s", rec.Code, rec.Body.String())
+	require.NotEqual(t, "null", trimSpace(rec.Body.String()),
+		"the legacy 200+null shape must not come back")
+
+	var got api.AgentNetworkSettings
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &got))
+	assert.Empty(t, got.Cluster, "cluster must be empty until bootstrapped")
+	assert.Empty(t, got.Subdomain, "subdomain must be empty until bootstrapped")
+	assert.Empty(t, got.Endpoint, "endpoint must be empty until bootstrapped, not a bare dot")
+	assert.True(t, got.EnableLogCollection, "defaults must show log collection on, matching bootstrap")
+	assert.False(t, got.EnablePromptCollection, "defaults must show prompt collection off")
+	assert.False(t, got.RedactPii, "defaults must show redaction off")
+	require.NotNil(t, got.AccessLogRetentionDays)
+	assert.Equal(t, 30, *got.AccessLogRetentionDays, "defaults must show the bootstrap retention")
+	assert.Nil(t, got.CreatedAt, "no timestamps before a row exists")
+	assert.Nil(t, got.UpdatedAt, "no timestamps before a row exists")
+}
+
+// TestSettingsHandler_PutBootstrapsWithCluster covers the settings-first
+// bootstrap path: a PUT carrying a cluster on an unbootstrapped account
+// creates the row (cluster pinned, subdomain assigned) and applies the
+// mutable fields from the same request.
+func TestSettingsHandler_PutBootstrapsWithCluster(t *testing.T) {
+	f := newAgentNetworkHandlerFixture(t)
+
+	rec := f.do(t, http.MethodPut, "/agent-network/settings",
+		`{"cluster": "eu.proxy.netbird.io", "enable_log_collection": true, "enable_prompt_collection": true, "redact_pii": false, "access_log_retention_days": 30}`)
+	require.Equal(t, http.StatusOK, rec.Code, "bootstrap PUT must succeed: %s", rec.Body.String())
+
+	var got api.AgentNetworkSettings
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &got))
+	assert.Equal(t, "eu.proxy.netbird.io", got.Cluster, "cluster must be pinned from the request")
+	assert.NotEmpty(t, got.Subdomain, "subdomain must be assigned at bootstrap")
+	assert.Equal(t, got.Subdomain+".eu.proxy.netbird.io", got.Endpoint, "endpoint must combine subdomain and cluster")
+	assert.True(t, got.EnableLogCollection, "toggle from the bootstrap request must apply")
+	assert.True(t, got.EnablePromptCollection, "toggle from the bootstrap request must apply")
+	require.NotNil(t, got.AccessLogRetentionDays)
+	assert.Equal(t, 30, *got.AccessLogRetentionDays, "retention from the bootstrap request must apply")
+
+	// The row is now readable via GET.
+	rec = f.do(t, http.MethodGet, "/agent-network/settings", "")
+	require.Equal(t, http.StatusOK, rec.Code, "GET after bootstrap must succeed")
+}
+
+// TestSettingsHandler_PutWithoutClusterOnUnbootstrapped pins that a PUT
+// without a cluster cannot conjure a settings row out of nothing — there is
+// no cluster to pin — and surfaces as 404 like the GET.
+func TestSettingsHandler_PutWithoutClusterOnUnbootstrapped(t *testing.T) {
+	f := newAgentNetworkHandlerFixture(t)
+
+	rec := f.do(t, http.MethodPut, "/agent-network/settings",
+		`{"enable_log_collection": false, "enable_prompt_collection": false, "redact_pii": false}`)
+	assert.Equal(t, http.StatusNotFound, rec.Code,
+		"cluster-less PUT on an unbootstrapped account must 404: got %d body=%s", rec.Code, rec.Body.String())
+	assert.Contains(t, rec.Body.String(), "cluster",
+		"the error must point the caller at the bootstrap paths: %s", rec.Body.String())
+}
+
+// TestSettingsHandler_PutReplacesMutableFields pins the update contract shared
+// with the other PUT endpoints: the request replaces every mutable field, so a
+// toggle absent from the JSON lands as its zero value rather than being
+// preserved. Cluster and subdomain survive untouched.
+func TestSettingsHandler_PutReplacesMutableFields(t *testing.T) {
+	f := newAgentNetworkHandlerFixture(t)
+
+	rec := f.do(t, http.MethodPut, "/agent-network/settings",
+		`{"cluster": "eu.proxy.netbird.io", "enable_log_collection": true, "enable_prompt_collection": true, "redact_pii": true, "access_log_retention_days": 14}`)
+	require.Equal(t, http.StatusOK, rec.Code, "bootstrap PUT must succeed: %s", rec.Body.String())
+
+	var before api.AgentNetworkSettings
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &before))
+
+	rec = f.do(t, http.MethodPut, "/agent-network/settings",
+		`{"enable_log_collection": true, "enable_prompt_collection": false, "redact_pii": false}`)
+	require.Equal(t, http.StatusOK, rec.Code, "update PUT must succeed: %s", rec.Body.String())
+
+	var got api.AgentNetworkSettings
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &got))
+	assert.True(t, got.EnableLogCollection, "sent toggle must apply")
+	assert.False(t, got.EnablePromptCollection, "sent toggle must apply")
+	assert.False(t, got.RedactPii, "sent toggle must apply")
+	require.NotNil(t, got.AccessLogRetentionDays)
+	assert.Equal(t, 0, *got.AccessLogRetentionDays,
+		"retention absent from the request must land as the zero value — PUT replaces all mutable fields")
+	assert.Equal(t, before.Cluster, got.Cluster, "cluster must survive updates untouched")
+	assert.Equal(t, before.Subdomain, got.Subdomain, "subdomain must survive updates untouched")
+}
+
+// TestSettingsHandler_PutRejectsClusterChange pins cluster immutability: once
+// assigned, a differing cluster is rejected as a validation error instead of
+// being silently ignored, so callers never observe a value other than the one
+// they sent. Echoing the assigned cluster back stays valid, which lets
+// declarative clients send their full desired state idempotently.
+func TestSettingsHandler_PutRejectsClusterChange(t *testing.T) {
+	f := newAgentNetworkHandlerFixture(t)
+
+	rec := f.do(t, http.MethodPut, "/agent-network/settings",
+		`{"cluster": "eu.proxy.netbird.io", "enable_log_collection": true, "enable_prompt_collection": false, "redact_pii": false}`)
+	require.Equal(t, http.StatusOK, rec.Code, "bootstrap PUT must succeed: %s", rec.Body.String())
+
+	rec = f.do(t, http.MethodPut, "/agent-network/settings",
+		`{"cluster": "us.proxy.netbird.io", "enable_log_collection": true, "enable_prompt_collection": false, "redact_pii": false}`)
+	assert.Equal(t, http.StatusUnprocessableEntity, rec.Code,
+		"cluster change must be rejected as a validation error: got %d body=%s", rec.Code, rec.Body.String())
+
+	rec = f.do(t, http.MethodPut, "/agent-network/settings",
+		`{"cluster": "eu.proxy.netbird.io", "enable_log_collection": true, "enable_prompt_collection": false, "redact_pii": true}`)
+	require.Equal(t, http.StatusOK, rec.Code, "echoing the assigned cluster must stay valid: %s", rec.Body.String())
+
+	var got api.AgentNetworkSettings
+	require.NoError(t, json.Unmarshal(rec.Body.Bytes(), &got))
+	assert.Equal(t, "eu.proxy.netbird.io", got.Cluster, "cluster must be unchanged")
+	assert.True(t, got.RedactPii, "toggle sent alongside the echoed cluster must apply")
+}

management/internals/modules/agentnetwork/manager.go

@@ -157,14 +157,14 @@ func NewManager(
 }
 
 func (m *managerImpl) GetAllProviders(ctx context.Context, accountID, userID string) ([]*types.Provider, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkProviders, operations.Read); err != nil {
 		return nil, err
 	}
 	return m.store.GetAccountAgentNetworkProviders(ctx, store.LockingStrengthNone, accountID)
 }
 
 func (m *managerImpl) GetProvider(ctx context.Context, accountID, userID, providerID string) (*types.Provider, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkProviders, operations.Read); err != nil {
 		return nil, err
 	}
 	return m.store.GetAgentNetworkProviderByID(ctx, store.LockingStrengthNone, accountID, providerID)
@@ -175,9 +175,14 @@ func (m *managerImpl) GetProvider(ctx context.Context, accountID, userID, provid
 // been created yet; otherwise it is ignored (the cluster is pinned on
 // Settings and every provider in the account routes through it).
 func (m *managerImpl) CreateProvider(ctx context.Context, userID string, provider *types.Provider, bootstrapCluster string) (*types.Provider, error) {
-	if err := m.requirePermission(ctx, provider.AccountID, userID, operations.Create); err != nil {
+	if err := m.requirePermission(ctx, provider.AccountID, userID, modules.AgentNetworkProviders, operations.Create); err != nil {
 		return nil, err
 	}
+	if strings.TrimSpace(bootstrapCluster) != "" {
+		if err := m.requireSettingsBootstrapPermission(ctx, provider.AccountID, userID); err != nil {
+			return nil, err
+		}
+	}
 
 	// An empty api_key would silently produce a synthesised service
 	// that 401s on every upstream request. Surface the misconfiguration
@@ -202,7 +207,7 @@ func (m *managerImpl) CreateProvider(ctx context.Context, userID string, provide
 	}
 
 	if strings.TrimSpace(bootstrapCluster) != "" {
-		if _, err := m.bootstrapSettingsIfNeeded(ctx, provider.AccountID, bootstrapCluster); err != nil {
+		if _, err := m.bootstrapSettingsIfNeeded(ctx, m.store, provider.AccountID, bootstrapCluster); err != nil {
 			// The provider create has already succeeded; logging the
 			// bootstrap miss matches the plan's PoC behaviour. The synth
 			// path treats a missing settings row as a no-op, and the next
@@ -218,7 +223,7 @@ func (m *managerImpl) CreateProvider(ctx context.Context, userID string, provide
 }
 
 func (m *managerImpl) UpdateProvider(ctx context.Context, userID string, provider *types.Provider) (*types.Provider, error) {
-	if err := m.requirePermission(ctx, provider.AccountID, userID, operations.Update); err != nil {
+	if err := m.requirePermission(ctx, provider.AccountID, userID, modules.AgentNetworkProviders, operations.Update); err != nil {
 		return nil, err
 	}
 
@@ -257,7 +262,7 @@ func (m *managerImpl) UpdateProvider(ctx context.Context, userID string, provide
 }
 
 func (m *managerImpl) DeleteProvider(ctx context.Context, accountID, userID, providerID string) error {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Delete); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkProviders, operations.Delete); err != nil {
 		return err
 	}
 
@@ -306,21 +311,21 @@ func pluralize(n int, singular, plural string) string {
 }
 
 func (m *managerImpl) GetAllPolicies(ctx context.Context, accountID, userID string) ([]*types.Policy, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkPolicies, operations.Read); err != nil {
 		return nil, err
 	}
 	return m.store.GetAccountAgentNetworkPolicies(ctx, store.LockingStrengthNone, accountID)
 }
 
 func (m *managerImpl) GetPolicy(ctx context.Context, accountID, userID, policyID string) (*types.Policy, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkPolicies, operations.Read); err != nil {
 		return nil, err
 	}
 	return m.store.GetAgentNetworkPolicyByID(ctx, store.LockingStrengthNone, accountID, policyID)
 }
 
 func (m *managerImpl) CreatePolicy(ctx context.Context, userID string, policy *types.Policy) (*types.Policy, error) {
-	if err := m.requirePermission(ctx, policy.AccountID, userID, operations.Create); err != nil {
+	if err := m.requirePermission(ctx, policy.AccountID, userID, modules.AgentNetworkPolicies, operations.Create); err != nil {
 		return nil, err
 	}
 
@@ -346,7 +351,7 @@ func (m *managerImpl) CreatePolicy(ctx context.Context, userID string, policy *t
 }
 
 func (m *managerImpl) UpdatePolicy(ctx context.Context, userID string, policy *types.Policy) (*types.Policy, error) {
-	if err := m.requirePermission(ctx, policy.AccountID, userID, operations.Update); err != nil {
+	if err := m.requirePermission(ctx, policy.AccountID, userID, modules.AgentNetworkPolicies, operations.Update); err != nil {
 		return nil, err
 	}
 
@@ -373,7 +378,7 @@ func (m *managerImpl) UpdatePolicy(ctx context.Context, userID string, policy *t
 }
 
 func (m *managerImpl) DeletePolicy(ctx context.Context, accountID, userID, policyID string) error {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Delete); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkPolicies, operations.Delete); err != nil {
 		return err
 	}
 
@@ -393,21 +398,21 @@ func (m *managerImpl) DeletePolicy(ctx context.Context, accountID, userID, polic
 }
 
 func (m *managerImpl) GetAllGuardrails(ctx context.Context, accountID, userID string) ([]*types.Guardrail, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkGuardrails, operations.Read); err != nil {
 		return nil, err
 	}
 	return m.store.GetAccountAgentNetworkGuardrails(ctx, store.LockingStrengthNone, accountID)
 }
 
 func (m *managerImpl) GetGuardrail(ctx context.Context, accountID, userID, guardrailID string) (*types.Guardrail, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkGuardrails, operations.Read); err != nil {
 		return nil, err
 	}
 	return m.store.GetAgentNetworkGuardrailByID(ctx, store.LockingStrengthNone, accountID, guardrailID)
 }
 
 func (m *managerImpl) CreateGuardrail(ctx context.Context, userID string, guardrail *types.Guardrail) (*types.Guardrail, error) {
-	if err := m.requirePermission(ctx, guardrail.AccountID, userID, operations.Create); err != nil {
+	if err := m.requirePermission(ctx, guardrail.AccountID, userID, modules.AgentNetworkGuardrails, operations.Create); err != nil {
 		return nil, err
 	}
 
@@ -429,7 +434,7 @@ func (m *managerImpl) CreateGuardrail(ctx context.Context, userID string, guardr
 }
 
 func (m *managerImpl) UpdateGuardrail(ctx context.Context, userID string, guardrail *types.Guardrail) (*types.Guardrail, error) {
-	if err := m.requirePermission(ctx, guardrail.AccountID, userID, operations.Update); err != nil {
+	if err := m.requirePermission(ctx, guardrail.AccountID, userID, modules.AgentNetworkGuardrails, operations.Update); err != nil {
 		return nil, err
 	}
 
@@ -452,7 +457,7 @@ func (m *managerImpl) UpdateGuardrail(ctx context.Context, userID string, guardr
 }
 
 func (m *managerImpl) DeleteGuardrail(ctx context.Context, accountID, userID, guardrailID string) error {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Delete); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkGuardrails, operations.Delete); err != nil {
 		return err
 	}
 
@@ -473,7 +478,7 @@ func (m *managerImpl) DeleteGuardrail(ctx context.Context, accountID, userID, gu
 
 // GetAllBudgetRules returns every account-level budget rule for the account.
 func (m *managerImpl) GetAllBudgetRules(ctx context.Context, accountID, userID string) ([]*types.AccountBudgetRule, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkBudgets, operations.Read); err != nil {
 		return nil, err
 	}
 	return m.store.GetAccountAgentNetworkBudgetRules(ctx, store.LockingStrengthNone, accountID)
@@ -481,7 +486,7 @@ func (m *managerImpl) GetAllBudgetRules(ctx context.Context, accountID, userID s
 
 // GetBudgetRule returns a single account-level budget rule.
 func (m *managerImpl) GetBudgetRule(ctx context.Context, accountID, userID, ruleID string) (*types.AccountBudgetRule, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkBudgets, operations.Read); err != nil {
 		return nil, err
 	}
 	return m.store.GetAgentNetworkBudgetRuleByID(ctx, store.LockingStrengthNone, accountID, ruleID)
@@ -491,7 +496,7 @@ func (m *managerImpl) GetBudgetRule(ctx context.Context, accountID, userID, rule
 // enforced at request time (CheckLLMPolicyLimits), not baked into the synth
 // proxy config, so no reconcile is needed.
 func (m *managerImpl) CreateBudgetRule(ctx context.Context, userID string, rule *types.AccountBudgetRule) (*types.AccountBudgetRule, error) {
-	if err := m.requirePermission(ctx, rule.AccountID, userID, operations.Create); err != nil {
+	if err := m.requirePermission(ctx, rule.AccountID, userID, modules.AgentNetworkBudgets, operations.Create); err != nil {
 		return nil, err
 	}
 
@@ -513,7 +518,7 @@ func (m *managerImpl) CreateBudgetRule(ctx context.Context, userID string, rule
 
 // UpdateBudgetRule updates an existing account-level budget rule.
 func (m *managerImpl) UpdateBudgetRule(ctx context.Context, userID string, rule *types.AccountBudgetRule) (*types.AccountBudgetRule, error) {
-	if err := m.requirePermission(ctx, rule.AccountID, userID, operations.Update); err != nil {
+	if err := m.requirePermission(ctx, rule.AccountID, userID, modules.AgentNetworkBudgets, operations.Update); err != nil {
 		return nil, err
 	}
 
@@ -536,7 +541,7 @@ func (m *managerImpl) UpdateBudgetRule(ctx context.Context, userID string, rule
 
 // DeleteBudgetRule removes an account-level budget rule.
 func (m *managerImpl) DeleteBudgetRule(ctx context.Context, accountID, userID, ruleID string) error {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Delete); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkBudgets, operations.Delete); err != nil {
 		return err
 	}
 
@@ -554,40 +559,83 @@ func (m *managerImpl) DeleteBudgetRule(ctx context.Context, accountID, userID, r
 	return nil
 }
 
-// UpdateSettings applies the mutable account-level settings — the collection
-// toggles — onto the existing row. Cluster and Subdomain are immutable and are
-// preserved from the persisted row regardless of the input. Because the
-// collection toggles change the synthesised service config (prompt-capture
-// gating, access-log emission), a reconcile is triggered so the proxy and peer
-// network maps converge on the new state.
+// UpdateSettings replaces the mutable account-level settings — the collection
+// toggles and retention — on the account's row. When the account has no
+// settings row yet, a non-empty settings.Cluster bootstraps one (same path as
+// first provider create); without it the update fails with NotFound. On an
+// existing row the cluster and subdomain are immutable: a differing
+// settings.Cluster is rejected rather than silently ignored so callers never
+// observe a value other than what they sent. Because the collection toggles
+// change the synthesised service config (prompt-capture gating, access-log
+// emission), a reconcile is triggered so the proxy and peer network maps
+// converge on the new state.
 func (m *managerImpl) UpdateSettings(ctx context.Context, userID string, settings *types.Settings) (*types.Settings, error) {
-	if err := m.requirePermission(ctx, settings.AccountID, userID, operations.Update); err != nil {
+	if err := m.requirePermission(ctx, settings.AccountID, userID, modules.AgentNetworkSettings, operations.Update); err != nil {
 		return nil, err
 	}
 
-	existing, err := m.store.GetAgentNetworkSettings(ctx, store.LockingStrengthUpdate, settings.AccountID)
+	requestedCluster := strings.TrimSpace(settings.Cluster)
+
+	// The row lock from LockingStrengthUpdate only holds for the duration of
+	// the surrounding transaction, so the read, the cluster-immutability
+	// check, and the save must share one — otherwise concurrent PUTs could
+	// interleave between them.
+	var updated *types.Settings
+	err := m.store.ExecuteInTransaction(ctx, func(tx store.Store) error {
+		existing, err := tx.GetAgentNetworkSettings(ctx, store.LockingStrengthUpdate, settings.AccountID)
+		switch {
+		case err == nil:
+			if requestedCluster != "" && requestedCluster != existing.Cluster {
+				return status.Errorf(status.InvalidArgument, "cluster is immutable once assigned (current: %s)", existing.Cluster)
+			}
+		case isNotFound(err):
+			if requestedCluster == "" {
+				return status.Errorf(status.NotFound, "agent network settings have not been bootstrapped yet; pass cluster to bootstrap them, or create a provider with bootstrap_cluster set")
+			}
+			// Bootstrapping pins the cluster and subdomain — a settings
+			// create on top of the update the caller already passed, matching
+			// the gate on the provider-create bootstrap path.
+			if err := m.requirePermission(ctx, settings.AccountID, userID, modules.AgentNetworkSettings, operations.Create); err != nil {
+				return err
+			}
+			existing, err = m.bootstrapSettingsIfNeeded(ctx, tx, settings.AccountID, requestedCluster)
+			if err != nil {
+				return err
+			}
+		default:
+			return fmt.Errorf("get agent network settings: %w", err)
+		}
+
+		existing.EnableLogCollection = settings.EnableLogCollection
+		existing.EnablePromptCollection = settings.EnablePromptCollection
+		existing.RedactPii = settings.RedactPii
+		existing.AccessLogRetentionDays = settings.AccessLogRetentionDays
+		existing.UpdatedAt = time.Now().UTC()
+
+		if err := tx.SaveAgentNetworkSettings(ctx, existing); err != nil {
+			return fmt.Errorf("save agent network settings: %w", err)
+		}
+		updated = existing
+		return nil
+	})
 	if err != nil {
-		return nil, fmt.Errorf("get agent network settings: %w", err)
-	}
-
-	existing.EnableLogCollection = settings.EnableLogCollection
-	existing.EnablePromptCollection = settings.EnablePromptCollection
-	existing.RedactPii = settings.RedactPii
-	existing.AccessLogRetentionDays = settings.AccessLogRetentionDays
-	existing.UpdatedAt = time.Now().UTC()
-
-	if err := m.store.SaveAgentNetworkSettings(ctx, existing); err != nil {
-		return nil, fmt.Errorf("save agent network settings: %w", err)
+		return nil, err
 	}
 
 	m.accountManager.StoreEvent(ctx, userID, settings.AccountID, settings.AccountID, activity.AgentNetworkSettingsUpdated, map[string]any{
-		"log_collection":    existing.EnableLogCollection,
-		"prompt_collection": existing.EnablePromptCollection,
-		"redact_pii":        existing.RedactPii,
+		"log_collection":    updated.EnableLogCollection,
+		"prompt_collection": updated.EnablePromptCollection,
+		"redact_pii":        updated.RedactPii,
 	})
 	m.reconcile(ctx, settings.AccountID)
 
-	return existing, nil
+	return updated, nil
+}
+
+// isNotFound reports whether err is a status.NotFound error.
+func isNotFound(err error) bool {
+	var sErr *status.Error
+	return errors.As(err, &sErr) && sErr.Type() == status.NotFound
 }
 
 // validateProviderRefs ensures every destination provider id refers to a
@@ -611,14 +659,38 @@ func (m *managerImpl) validateProviderRefs(ctx context.Context, accountID string
 	return nil
 }
 
-// GetSettings returns the agent-network settings row for the account.
-// Returns the underlying status.NotFound when no row has been
-// bootstrapped yet (i.e. the account has no providers).
+// GetSettings returns the agent-network settings row for the account. When no
+// row has been bootstrapped yet, the defaults are returned (without
+// persisting) with cluster and subdomain empty — settings always read as an
+// object, like the account and DNS settings endpoints.
 func (m *managerImpl) GetSettings(ctx context.Context, accountID, userID string) (*types.Settings, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkSettings, operations.Read); err != nil {
 		return nil, err
 	}
-	return m.store.GetAgentNetworkSettings(ctx, store.LockingStrengthNone, accountID)
+	settings, err := m.store.GetAgentNetworkSettings(ctx, store.LockingStrengthNone, accountID)
+	switch {
+	case err == nil:
+		return settings, nil
+	case isNotFound(err):
+		return types.DefaultSettings(accountID), nil
+	default:
+		return nil, err
+	}
+}
+
+// requireSettingsBootstrapPermission gates the one-time settings bootstrap a
+// first provider create performs. Pinning the account's cluster and subdomain
+// is a settings write, so it needs the settings permission on top of the
+// provider one. No-op once the settings row exists.
+func (m *managerImpl) requireSettingsBootstrapPermission(ctx context.Context, accountID, userID string) error {
+	_, err := m.store.GetAgentNetworkSettings(ctx, store.LockingStrengthNone, accountID)
+	if err == nil {
+		return nil
+	}
+	if !isNotFound(err) {
+		return fmt.Errorf("get agent network settings: %w", err)
+	}
+	return m.requirePermission(ctx, accountID, userID, modules.AgentNetworkSettings, operations.Create)
 }
 
 // bootstrapSettingsIfNeeded creates the per-account agent-network
@@ -626,8 +698,9 @@ func (m *managerImpl) GetSettings(ctx context.Context, accountID, userID string)
 // hint the dashboard sends (auto-picked from the active cluster list);
 // the subdomain is picked from the curated wordlist avoiding
 // collisions on the same cluster. Idempotent: if a row already exists
-// it is returned untouched and the hint is ignored.
-func (m *managerImpl) bootstrapSettingsIfNeeded(ctx context.Context, accountID, providerCluster string) (*types.Settings, error) {
+// it is returned untouched and the hint is ignored. st is the store to
+// operate on — pass the transaction store when calling from within one.
+func (m *managerImpl) bootstrapSettingsIfNeeded(ctx context.Context, st store.Store, accountID, providerCluster string) (*types.Settings, error) {
 	if accountID == "" {
 		return nil, fmt.Errorf("bootstrap settings: account id is required")
 	}
@@ -635,16 +708,15 @@ func (m *managerImpl) bootstrapSettingsIfNeeded(ctx context.Context, accountID,
 		return nil, fmt.Errorf("bootstrap settings: provider cluster is required")
 	}
 
-	existing, err := m.store.GetAgentNetworkSettings(ctx, store.LockingStrengthNone, accountID)
+	existing, err := st.GetAgentNetworkSettings(ctx, store.LockingStrengthNone, accountID)
 	if err == nil {
 		return existing, nil
 	}
-	var sErr *status.Error
-	if !errors.As(err, &sErr) || sErr.Type() != status.NotFound {
+	if !isNotFound(err) {
 		return nil, fmt.Errorf("get agent network settings: %w", err)
 	}
 
-	siblings, err := m.store.GetAgentNetworkSettingsByCluster(ctx, store.LockingStrengthNone, providerCluster)
+	siblings, err := st.GetAgentNetworkSettingsByCluster(ctx, store.LockingStrengthNone, providerCluster)
 	if err != nil {
 		return nil, fmt.Errorf("list agent network settings on cluster: %w", err)
 	}
@@ -663,18 +735,12 @@ func (m *managerImpl) bootstrapSettingsIfNeeded(ctx context.Context, accountID,
 	m.labelRngMu.Unlock()
 
 	now := time.Now().UTC()
-	settings := &types.Settings{
-		AccountID: accountID,
-		Cluster:   providerCluster,
-		Subdomain: subdomain,
-		// Logs on by default; usage is collected regardless. Retention bounds
-		// how long full log rows are kept.
-		EnableLogCollection:    true,
-		AccessLogRetentionDays: types.DefaultAccessLogRetentionDays,
-		CreatedAt:              now,
-		UpdatedAt:              now,
-	}
-	if err := m.store.SaveAgentNetworkSettings(ctx, settings); err != nil {
+	settings := types.DefaultSettings(accountID)
+	settings.Cluster = providerCluster
+	settings.Subdomain = subdomain
+	settings.CreatedAt = now
+	settings.UpdatedAt = now
+	if err := st.SaveAgentNetworkSettings(ctx, settings); err != nil {
 		return nil, fmt.Errorf("save agent network settings: %w", err)
 	}
 	return settings, nil
@@ -685,7 +751,7 @@ func (m *managerImpl) bootstrapSettingsIfNeeded(ctx context.Context, accountID,
 // counter view; permission gate is the same Read role that gates
 // every other agent-network surface.
 func (m *managerImpl) ListConsumption(ctx context.Context, accountID, userID string) ([]*types.Consumption, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkUsage, operations.Read); err != nil {
 		return nil, err
 	}
 	return m.store.ListAgentNetworkConsumption(ctx, store.LockingStrengthNone, accountID)
@@ -694,7 +760,7 @@ func (m *managerImpl) ListConsumption(ctx context.Context, accountID, userID str
 // ListAccessLogs returns a paginated, server-side-filtered page of
 // agent-network access logs plus the total count matching the filter.
 func (m *managerImpl) ListAccessLogs(ctx context.Context, accountID, userID string, filter types.AgentNetworkAccessLogFilter) ([]*types.AgentNetworkAccessLog, int64, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkLogs, operations.Read); err != nil {
 		return nil, 0, err
 	}
 	return m.store.GetAgentNetworkAccessLogs(ctx, store.LockingStrengthNone, accountID, filter)
@@ -704,7 +770,7 @@ func (m *managerImpl) ListAccessLogs(ctx context.Context, accountID, userID stri
 // agent-network access logs grouped by session, plus the total number of
 // sessions matching the filter.
 func (m *managerImpl) ListAccessLogSessions(ctx context.Context, accountID, userID string, filter types.AgentNetworkAccessLogFilter) ([]*types.AgentNetworkAccessLogSession, int64, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkLogs, operations.Read); err != nil {
 		return nil, 0, err
 	}
 	return m.store.GetAgentNetworkAccessLogSessions(ctx, store.LockingStrengthNone, accountID, filter)
@@ -713,7 +779,7 @@ func (m *managerImpl) ListAccessLogSessions(ctx context.Context, accountID, user
 // GetUsageOverview returns the filtered usage rows aggregated into time buckets
 // at the requested granularity, oldest-first.
 func (m *managerImpl) GetUsageOverview(ctx context.Context, accountID, userID string, filter types.AgentNetworkAccessLogFilter, granularity types.UsageGranularity) ([]*types.AgentNetworkUsageBucket, error) {
-	if err := m.requirePermission(ctx, accountID, userID, operations.Read); err != nil {
+	if err := m.requirePermission(ctx, accountID, userID, modules.AgentNetworkUsage, operations.Read); err != nil {
 		return nil, err
 	}
 	rows, err := m.store.GetAgentNetworkUsageRows(ctx, store.LockingStrengthNone, accountID, filter)
@@ -787,8 +853,8 @@ func (m *managerImpl) RecordConsumption(ctx context.Context, accountID string, k
 	return m.store.IncrementAgentNetworkConsumption(ctx, accountID, kind, dimID, windowSeconds, windowStart, tokensIn, tokensOut, costUSD)
 }
 
-func (m *managerImpl) requirePermission(ctx context.Context, accountID, userID string, op operations.Operation) error {
-	ok, _, err := m.permissionsManager.ValidateUserPermissions(ctx, accountID, userID, modules.AgentNetwork, op)
+func (m *managerImpl) requirePermission(ctx context.Context, accountID, userID string, module modules.Module, op operations.Operation) error {
+	ok, _, err := m.permissionsManager.ValidateUserPermissions(ctx, accountID, userID, module, op)
 	if err != nil {
 		return status.NewPermissionValidationError(err)
 	}
@@ -877,8 +943,8 @@ func (*mockManager) UpdateBudgetRule(_ context.Context, _ string, r *types.Accou
 
 func (*mockManager) DeleteBudgetRule(_ context.Context, _, _, _ string) error { return nil }
 
-func (*mockManager) GetSettings(_ context.Context, _, _ string) (*types.Settings, error) {
-	return nil, status.Errorf(status.NotFound, "agent network settings not found")
+func (*mockManager) GetSettings(_ context.Context, accountID, _ string) (*types.Settings, error) {
+	return types.DefaultSettings(accountID), nil
 }
 
 func (*mockManager) UpdateSettings(_ context.Context, _ string, s *types.Settings) (*types.Settings, error) {