Merge branch 'feature/android-peer-detail' into android/gui-integration

The partial-failure test only placed the unknown ID after the valid
exit node, so it never verified that selection processing continues
past a leading invalid ID. Run the same assertions for both orderings.

.github/workflows/agent-network-e2e.yml

@@ -5,6 +5,13 @@ on:
   schedule:
     - cron: "0 3 * * *"
   workflow_dispatch:
+    inputs:
+      bedrock_model:
+        description: >-
+          Bedrock inference-profile id to drive the matrix with, exactly as
+          AWS issues it. Leave empty for the Sonnet 4.6 default.
+        required: false
+        default: ""
 
 concurrency:
   group: ${{ github.workflow }}-${{ github.ref }}
@@ -51,6 +58,9 @@ jobs:
           # token (and URL, for gateways) is unset, so partial coverage is fine.
           OPENAI_TOKEN: ${{ secrets.E2E_OPENAI_TOKEN }}
           ANTHROPIC_TOKEN: ${{ secrets.E2E_ANTHROPIC_TOKEN }}
+          # Moonshot AI platform key (platform.kimi.ai); drives both Kimi wire
+          # shapes (OpenAI /v1 and Anthropic /anthropic) through kimi_api.
+          KIMI_TOKEN: ${{ secrets.E2E_KIMI_TOKEN }}
           VERCEL_URL: ${{ secrets.E2E_VERCEL_URL }}
           VERCEL_TOKEN: ${{ secrets.E2E_VERCEL_TOKEN }}
           OPENROUTER_URL: ${{ secrets.E2E_OPENROUTER_URL }}
@@ -59,6 +69,8 @@ jobs:
           CLOUDFLARE_TOKEN: ${{ secrets.E2E_CLOUDFLARE_TOKEN }}
           AWS_BEARER_TOKEN_BEDROCK: ${{ secrets.E2E_AWS_BEARER_TOKEN_BEDROCK }}
           AWS_REGION: ${{ secrets.E2E_AWS_REGION }}
+          # Bedrock model override: dispatch input wins, then the repo variable, else the test default.
+          AWS_BEDROCK_MODEL: ${{ inputs.bedrock_model || vars.E2E_AWS_BEDROCK_MODEL }}
           # Vertex (Anthropic-on-Vertex): SA + project required; region defaults
           # to "global", model to a pinned claude snapshot.
           GOOGLE_VERTEX_SA_BASE64: ${{ secrets.E2E_GOOGLE_VERTEX_SA_BASE64 }}

.github/workflows/frontend-ui.yml

@@ -86,7 +86,7 @@ jobs:
             ${{ runner.os }}-pnpm-
 
       - name: Install dependencies
-        run: pnpm install --frozen-lockfile
+        run: pnpm install --frozen-lockfile --ignore-scripts
 
       - name: Generate Wails bindings
         run: pnpm run bindings

.github/workflows/golangci-lint.yml

@@ -45,7 +45,7 @@ jobs:
             display_name: Linux
     name: ${{ matrix.display_name }}
     runs-on: ${{ matrix.os }}
-    timeout-minutes: 15
+    timeout-minutes: 25
     steps:
       - name: Checkout code
         uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
@@ -79,4 +79,4 @@ jobs:
           skip-cache: true
           skip-save-cache: true
           cache-invalidation-interval: 0
-          args: --timeout=12m
+          args: --timeout=20m

.goreleaser.yaml

@@ -273,8 +273,8 @@ dockers_v2:
        - netbirdio/netbird
        - ghcr.io/netbirdio/netbird
      tags:
-       - "v{{ .Version }}-rootless"
-       - "{{ if eq .Env.SKIP_PUBLISH \"false\" }}latest{{ end }}"
+       - "{{ .Version }}-rootless"
+       - "{{ if eq .Env.SKIP_PUBLISH \"false\" }}rootless-latest{{ end }}"
      dockerfile: client/Dockerfile-rootless
      extra_files:
        - client/netbird-entrypoint.sh

client/android/client.go

@@ -7,6 +7,7 @@ import (
 	"fmt"
 	"os"
 	"slices"
+	"strings"
 	"sync"
 	"time"
 
@@ -299,6 +300,13 @@ func (c *Client) SetInfoLogLevel() {
 // PeersList return with the list of the PeerInfos
 func (c *Client) PeersList() *PeerInfoArray {
 
+	// The recorder only caches transfer counters and handshake times; nothing
+	// refreshes them on its own, so without this they read as zero. The desktop
+	// daemon does the same before serving a full peer status.
+	if err := c.recorder.RefreshWireGuardStats(); err != nil {
+		log.Debugf("failed to refresh WireGuard stats: %v", err)
+	}
+
 	fullStatus := c.recorder.GetFullStatus()
 
 	peerInfos := make([]PeerInfo, len(fullStatus.Peers))
@@ -309,6 +317,20 @@ func (c *Client) PeersList() *PeerInfoArray {
 			FQDN:       p.FQDN,
 			ConnStatus: int(p.ConnStatus),
 			Routes:     PeerRoutes{routes: maps.Keys(p.GetRoutes())},
+
+			PubKey:                     p.PubKey,
+			Latency:                    formatDuration(p.Latency),
+			LatencyMs:                  p.Latency.Milliseconds(),
+			BytesRx:                    p.BytesRx,
+			BytesTx:                    p.BytesTx,
+			ConnStatusUpdate:           formatTime(p.ConnStatusUpdate),
+			Relayed:                    p.Relayed,
+			RosenpassEnabled:           p.RosenpassEnabled,
+			LastWireguardHandshake:     formatTime(p.LastWireguardHandshake),
+			LocalIceCandidateType:      p.LocalIceCandidateType,
+			RemoteIceCandidateType:     p.RemoteIceCandidateType,
+			LocalIceCandidateEndpoint:  p.LocalIceCandidateEndpoint,
+			RemoteIceCandidateEndpoint: p.RemoteIceCandidateEndpoint,
 		}
 		peerInfos[n] = pi
 	}
@@ -439,10 +461,6 @@ func (c *Client) RemoveConnectionListener() {
 	c.recorder.RemoveConnectionListener()
 }
 
-func (c *Client) toggleRoute(command routeCommand) error {
-	return command.toggleRoute()
-}
-
 func (c *Client) getRouteManager() (routemanager.Manager, error) {
 	client := c.getConnectClient()
 	if client == nil {
@@ -462,22 +480,22 @@ func (c *Client) getRouteManager() (routemanager.Manager, error) {
 	return manager, nil
 }
 
-func (c *Client) SelectRoute(route string) error {
+func (c *Client) SelectRoute(id string) error {
 	manager, err := c.getRouteManager()
 	if err != nil {
 		return err
 	}
 
-	return c.toggleRoute(selectRouteCommand{route: route, manager: manager})
+	return manager.SelectRoutes([]route.NetID{route.NetID(id)}, true)
 }
 
-func (c *Client) DeselectRoute(route string) error {
+func (c *Client) DeselectRoute(id string) error {
 	manager, err := c.getRouteManager()
 	if err != nil {
 		return err
 	}
 
-	return c.toggleRoute(deselectRouteCommand{route: route, manager: manager})
+	return manager.DeselectRoutes([]route.NetID{route.NetID(id)})
 }
 
 // getNetworkDomainsFromRoute extracts domains from a route and enriches each domain
@@ -512,3 +530,28 @@ func exportEnvList(list *EnvList) {
 		}
 	}
 }
+
+// formatDuration renders a duration for display, trimming the fractional part
+// to two digits so latencies read as "12.34ms" rather than "12.345678ms".
+func formatDuration(d time.Duration) string {
+	ds := d.String()
+	dotIndex := strings.Index(ds, ".")
+	if dotIndex == -1 {
+		return ds
+	}
+
+	endIndex := min(dotIndex+3, len(ds))
+
+	// Skip the remaining digits so only the unit suffix is appended back.
+	unitStart := endIndex
+	for unitStart < len(ds) && ds[unitStart] >= '0' && ds[unitStart] <= '9' {
+		unitStart++
+	}
+	return ds[:endIndex] + ds[unitStart:]
+}
+
+// formatTime renders a timestamp in UTC using a fixed layout. The zero time is
+// passed through as-is so the UI can recognise it and show "never" instead.
+func formatTime(t time.Time) string {
+	return t.UTC().Format("2006-01-02 15:04:05")
+}

client/android/peer_notifier.go

@@ -12,12 +12,30 @@ const (
 )
 
 // PeerInfo describe information about the peers. It designed for the UI usage
+//
+// The fields below ConnStatus back the peer detail screen. Durations and times
+// are pre-formatted into strings so the UI does not have to know Go's layouts;
+// Latency is additionally exposed as LatencyMs for colour coding.
 type PeerInfo struct {
 	IP         string
 	IPv6       string
 	FQDN       string
 	ConnStatus int
 	Routes     PeerRoutes
+
+	PubKey                     string
+	Latency                    string
+	LatencyMs                  int64
+	BytesRx                    int64
+	BytesTx                    int64
+	ConnStatusUpdate           string
+	Relayed                    bool
+	RosenpassEnabled           bool
+	LastWireguardHandshake     string
+	LocalIceCandidateType      string
+	RemoteIceCandidateType     string
+	LocalIceCandidateEndpoint  string
+	RemoteIceCandidateEndpoint string
 }
 
 func (p *PeerInfo) GetPeerRoutes() *PeerRoutes {

client/android/profile_manager.go

@@ -189,6 +189,19 @@ func (pm *ProfileManager) LogoutProfile(id string) error {
 	return nil
 }
 
+// RenameProfile changes a profile's display name. The profile ID, and therefore
+// its on-disk filename, is left untouched: only the "name" field of the config
+// is rewritten. This works for the default profile too, whose config lives in
+// netbird.cfg rather than under profiles/.
+func (pm *ProfileManager) RenameProfile(id string, newName string) error {
+	if err := pm.serviceMgr.RenameProfile(profilemanager.ID(id), androidUsername, newName); err != nil {
+		return fmt.Errorf("failed to rename profile: %w", err)
+	}
+
+	log.Infof("renamed profile %s to: %s", id, newName)
+	return nil
+}
+
 // RemoveProfile deletes a profile
 func (pm *ProfileManager) RemoveProfile(id string) error {
 	// Use ServiceManager (removes profile from profiles/ directory)

client/android/route_command.go

@@ -1,70 +0,0 @@
-//go:build android
-
-package android
-
-import (
-	"fmt"
-
-	log "github.com/sirupsen/logrus"
-	"golang.org/x/exp/maps"
-
-	"github.com/netbirdio/netbird/client/internal/routemanager"
-	"github.com/netbirdio/netbird/route"
-)
-
-func executeRouteToggle(id string, manager routemanager.Manager,
-	operationName string,
-	routeOperation func(routes []route.NetID, allRoutes []route.NetID) error) error {
-	netID := route.NetID(id)
-	routes := []route.NetID{netID}
-
-	routesMap := manager.GetClientRoutesWithNetID()
-	routes = route.ExpandV6ExitPairs(routes, routesMap)
-
-	log.Debugf("%s with ids: %v", operationName, routes)
-
-	if err := routeOperation(routes, maps.Keys(routesMap)); err != nil {
-		log.Debugf("error when %s: %s", operationName, err)
-		return fmt.Errorf("error %s: %w", operationName, err)
-	}
-
-	manager.TriggerSelection(manager.GetClientRoutes())
-
-	return nil
-}
-
-type routeCommand interface {
-	toggleRoute() error
-}
-
-type selectRouteCommand struct {
-	route   string
-	manager routemanager.Manager
-}
-
-func (s selectRouteCommand) toggleRoute() error {
-	routeSelector := s.manager.GetRouteSelector()
-	if routeSelector == nil {
-		return fmt.Errorf("no route selector available")
-	}
-
-	routeOperation := func(routes []route.NetID, allRoutes []route.NetID) error {
-		return routeSelector.SelectRoutes(routes, true, allRoutes)
-	}
-
-	return executeRouteToggle(s.route, s.manager, "selecting route", routeOperation)
-}
-
-type deselectRouteCommand struct {
-	route   string
-	manager routemanager.Manager
-}
-
-func (d deselectRouteCommand) toggleRoute() error {
-	routeSelector := d.manager.GetRouteSelector()
-	if routeSelector == nil {
-		return fmt.Errorf("no route selector available")
-	}
-
-	return executeRouteToggle(d.route, d.manager, "deselecting route", routeSelector.DeselectRoutes)
-}

client/iface/iface_test.go

@@ -464,6 +464,8 @@ func Test_RemovePeer(t *testing.T) {
 }
 
 func Test_ConnectPeers(t *testing.T) {
+	t.Setenv("NB_DISABLE_EBPF_WG_PROXY", "true")
+
 	peer1ifaceName := fmt.Sprintf("utun%d", WgIntNumber+400)
 	peer1wgIP := netip.MustParsePrefix("10.99.99.17/30")
 	peer1Key, _ := wgtypes.GeneratePrivateKey()
@@ -505,12 +507,8 @@ func Test_ConnectPeers(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	localIP, err := getLocalIP()
-	if err != nil {
-		t.Fatal(err)
-	}
-
-	peer1endpoint, err := net.ResolveUDPAddr("udp", fmt.Sprintf("%s:%d", localIP, peer1wgPort))
+	localIP1 := "127.0.0.1"
+	peer1endpoint, err := net.ResolveUDPAddr("udp", fmt.Sprintf("%s:%d", localIP1, peer1wgPort))
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -546,7 +544,8 @@ func Test_ConnectPeers(t *testing.T) {
 		t.Fatal(err)
 	}
 
-	peer2endpoint, err := net.ResolveUDPAddr("udp", fmt.Sprintf("%s:%d", localIP, peer2wgPort))
+	localIP2 := "127.0.0.1"
+	peer2endpoint, err := net.ResolveUDPAddr("udp", fmt.Sprintf("%s:%d", localIP2, peer2wgPort))
 	if err != nil {
 		t.Fatal(err)
 	}
@@ -569,17 +568,17 @@ func Test_ConnectPeers(t *testing.T) {
 	if err != nil {
 		t.Fatal(err)
 	}
-	// todo: investigate why in some tests execution we need 30s
+	// The peers use userspace WireGuard (stdnet transport). A tight busy-loop
+	// here starves the wireguard-go goroutines that process the handshake, so
+	// poll on a ticker instead and yield the CPU between checks. WireGuard also
+	// only retries a lost handshake initiation every REKEY_TIMEOUT (5s), which
+	// is why the overall wait can occasionally stretch to tens of seconds.
 	timeout := 30 * time.Second
 	timeoutChannel := time.After(timeout)
+	ticker := time.NewTicker(500 * time.Millisecond)
+	defer ticker.Stop()
 
 	for {
-		select {
-		case <-timeoutChannel:
-			t.Fatalf("waiting for peer handshake timeout after %s", timeout.String())
-		default:
-		}
-
 		peer, gpErr := getPeer(peer1ifaceName, peer2Key.PublicKey().String())
 		if gpErr != nil {
 			t.Fatal(gpErr)
@@ -588,6 +587,12 @@ func Test_ConnectPeers(t *testing.T) {
 			t.Log("peers successfully handshake")
 			break
 		}
+
+		select {
+		case <-timeoutChannel:
+			t.Fatalf("waiting for peer handshake timeout after %s", timeout.String())
+		case <-ticker.C:
+		}
 	}
 
 }
@@ -615,28 +620,3 @@ func getPeer(ifaceName, peerPubKey string) (wgtypes.Peer, error) {
 	}
 	return wgtypes.Peer{}, fmt.Errorf("peer not found")
 }
-
-func getLocalIP() (string, error) {
-	// Get all interfaces
-	addrs, err := net.InterfaceAddrs()
-	if err != nil {
-		return "", err
-	}
-
-	for _, addr := range addrs {
-		ipNet, ok := addr.(*net.IPNet)
-		if !ok {
-			continue
-		}
-		if ipNet.IP.IsLoopback() {
-			continue
-		}
-
-		if ipNet.IP.To4() == nil {
-			continue
-		}
-		return ipNet.IP.String(), nil
-	}
-
-	return "", fmt.Errorf("no local IP found")
-}

client/internal/auth/auth.go

@@ -351,6 +351,7 @@ func (a *Auth) setSystemInfoFlags(info *system.Info) {
 		a.config.BlockLANAccess,
 		a.config.BlockInbound,
 		a.config.DisableIPv6,
+		a.config.SyncMessageVersion,
 		a.config.EnableSSHRoot,
 		a.config.EnableSSHSFTP,
 		a.config.EnableSSHLocalPortForwarding,

client/internal/auth/sessionwatch/watcher.go

@@ -24,11 +24,7 @@ import (
 )
 
 const (
-	// Skew tolerates a small clock difference between the management
-	// server and this peer before treating a deadline as "in the past".
-	// Slightly above typical NTP drift; tight enough that the UI doesn't
-	// paint a stale expiry as if it were valid.
-	Skew = 30 * time.Second
+	maxPastHorizon = 30 * 24 * time.Hour
 
 	// maxDeadlineHorizon caps how far in the future an accepted deadline
 	// can sit. A timestamp beyond this is almost certainly a protocol
@@ -57,7 +53,7 @@ var (
 	ErrDeadlineTooFarFuture = errors.New("session deadline too far in the future")
 
 	// ErrDeadlineInPast is returned by Update when the supplied deadline
-	// is more than Skew in the past.
+	// is more than maxPastHorizon in the past.
 	ErrDeadlineInPast = errors.New("session deadline in the past")
 )
 
@@ -66,15 +62,14 @@ var (
 // for deadline change/clear, PublishEvent for the two warnings); tests pass
 // a fake recorder so the same surface is observable without an engine.
 //
-// The watcher is the single owner of the deadline propagated to the
-// recorder: every set, clear, sanity-check rejection and Close routes the
-// value through SetSessionExpiresAt, so the SubscribeStatus snapshot the UI
-// reads can never drift from the watcher's timer state. (SetSessionExpiresAt
-// fans out its own state-change notification, so no separate notify is
-// needed.) The recorder is server-scoped and outlives this engine-scoped
-// watcher — without the Close-time clear a teardown (Down, or the Down+Up of
-// a profile switch) would leave the next session showing the previous one's
-// stale "expires in" value.
+// While the watcher runs, it owns the deadline propagated to the recorder:
+// every set, clear and sanity-check rejection routes the value through
+// SetSessionExpiresAt, so the SubscribeStatus snapshot the UI reads can
+// never drift from the watcher's timer state. (SetSessionExpiresAt fans
+// out its own state-change notification, so no separate notify is needed.)
+// The recorder is server-scoped and outlives this engine-scoped watcher;
+// Close deliberately leaves the recorder value in place so transient engine
+// restarts don't blank it — the client run loop clears it on real teardown.
 //
 // PublishEvent's signature mirrors peer.Status.PublishEvent: the watcher
 // composes the metadata internally so the wire format (MetaSession*) is
@@ -135,10 +130,13 @@ func NewWithLeads(lead, final time.Duration, recorder StatusRecorder) *Watcher {
 // was disabled).
 //
 // Same-value updates are no-ops. A different non-zero value cancels any
-// pending timer, resets the "already fired" guard, and arms a new one.
+// pending timer, resets the "already fired" guards, and — when the
+// deadline lies in the future — arms fresh warning timers. A deadline
+// already in the past (within maxPastHorizon) is recorded as-is with no
+// timers: the session has expired and consumers render it that way.
 //
 // Returns one of the sentinel Err* values when the deadline fails the
-// sanity checks (pre-epoch, far future, or in the past beyond Skew).
+// sanity checks (pre-epoch, far future, or past beyond maxPastHorizon).
 // In every error case the watcher first clears its state so it stays
 // consistent with what the caller will push into its other sinks (e.g.
 // applySessionDeadline forces a zero deadline into the status recorder
@@ -163,7 +161,7 @@ func (w *Watcher) Update(deadline time.Time) error {
 	case deadline.After(now.Add(maxDeadlineHorizon)):
 		w.clearLocked()
 		return fmt.Errorf("%w: %v", ErrDeadlineTooFarFuture, deadline)
-	case deadline.Before(now.Add(-Skew)):
+	case deadline.Before(now.Add(-maxPastHorizon)):
 		w.clearLocked()
 		return fmt.Errorf("%w: %v (now=%v)", ErrDeadlineInPast, deadline, now)
 	}
@@ -183,7 +181,9 @@ func (w *Watcher) Update(deadline time.Time) error {
 	w.finalFiredAt = time.Time{}
 	w.dismissedAt = time.Time{}
 
-	w.armTimerLocked(deadline)
+	if deadline.After(now) {
+		w.armTimerLocked(deadline)
+	}
 	recorder := w.recorder
 	w.mu.Unlock()
 	if recorder != nil {
@@ -227,30 +227,25 @@ func (w *Watcher) Dismiss() {
 	log.Infof("auth session final-warning dismissed for deadline %s", w.current.Format(time.RFC3339))
 }
 
-// Close stops any pending timer and drops the deadline on the status
-// recorder. Update calls after Close are ignored. Clearing the recorder
-// here is what keeps a teardown (Down, or the Down+Up of a profile switch)
-// from leaving the next session showing this one's stale "expires in"
-// value — the recorder is server-scoped and outlives this engine-scoped
-// watcher, so nothing else drops the anchor on teardown.
+// Close stops any pending timer. Update calls after Close are ignored.
+// The recorder keeps its deadline: the watcher is engine-scoped and closes
+// on every engine restart (network change, sleep/wake, stream errors)
+// while the SSO deadline stays valid across those, so clearing here would
+// blank the UI's "expires in" row on every transient reconnect. The
+// client run loop clears the server-scoped recorder when it exits for
+// real (Down, profile switch, permanent login failure).
 func (w *Watcher) Close() {
 	w.mu.Lock()
+	defer w.mu.Unlock()
 	if w.closed {
-		w.mu.Unlock()
 		return
 	}
 	w.closed = true
 	w.stopTimerLocked()
-	hadDeadline := !w.current.IsZero()
 	w.current = time.Time{}
 	w.firedAt = time.Time{}
 	w.finalFiredAt = time.Time{}
 	w.dismissedAt = time.Time{}
-	recorder := w.recorder
-	w.mu.Unlock()
-	if recorder != nil && hadDeadline {
-		recorder.SetSessionExpiresAt(time.Time{})
-	}
 }
 
 // clearLocked drops the tracked deadline and notifies the recorder so

client/internal/auth/sessionwatch/watcher_test.go

@@ -224,11 +224,13 @@ func TestNewDeadlineCancelsPriorTimer(t *testing.T) {
 
 func TestRefreshAfterFireArmsNewWarning(t *testing.T) {
 	r := &fakeRecorder{}
-	lead := 30 * time.Millisecond
+	lead := 150 * time.Millisecond
 	w := newWatcher(lead, r)
 	defer w.Close()
 
-	first := time.Now().Add(50 * time.Millisecond)
+	// Warning fires ~20ms in; the deadline itself stays 150ms away so the
+	// replacement below lands well before it.
+	first := time.Now().Add(170 * time.Millisecond)
 	_ = w.Update(first)
 
 	// Wait for stateChange + warning of the first cycle.
@@ -306,7 +308,29 @@ func TestUpdateRejectsTooFarFuture(t *testing.T) {
 	}
 }
 
-func TestUpdateInPastClearsDeadline(t *testing.T) {
+func TestUpdateRecentPastRecordedAsExpired(t *testing.T) {
+	r := &fakeRecorder{}
+	w := newWatcher(50*time.Millisecond, r)
+	defer w.Close()
+
+	d := time.Now().Add(-1 * time.Hour)
+	if err := w.Update(d); err != nil {
+		t.Fatalf("recent-past Update should succeed, got %v", err)
+	}
+	if !w.Deadline().Equal(d) {
+		t.Fatalf("expected deadline to be recorded, got %v want %v", w.Deadline(), d)
+	}
+	if got := r.deadline(); !got.Equal(d) {
+		t.Fatalf("recorder deadline = %v, want %v", got, d)
+	}
+
+	time.Sleep(80 * time.Millisecond)
+	if n := countWhere(r.snapshot(), func(e event) bool { return e.kind == publish }); n != 0 {
+		t.Fatalf("no warning events may fire for an already-past deadline, got %+v", r.snapshot())
+	}
+}
+
+func TestUpdateAncientPastRejected(t *testing.T) {
 	r := &fakeRecorder{}
 	w := newWatcher(50*time.Millisecond, r)
 	defer w.Close()
@@ -318,12 +342,12 @@ func TestUpdateInPastClearsDeadline(t *testing.T) {
 	// Drain the stateChange from the seed.
 	waitForEvents(t, r, 1)
 
-	err := w.Update(time.Now().Add(-1 * time.Hour))
+	err := w.Update(time.Now().Add(-31 * 24 * time.Hour))
 	if !errors.Is(err, ErrDeadlineInPast) {
 		t.Fatalf("want ErrDeadlineInPast, got %v", err)
 	}
 	if !w.Deadline().IsZero() {
-		t.Fatalf("in-past update must clear the deadline, got %v", w.Deadline())
+		t.Fatalf("rejected ancient-past update must clear the deadline, got %v", w.Deadline())
 	}
 	events := waitForEvents(t, r, 2)
 	if events[1].kind != stateChange {
@@ -331,39 +355,25 @@ func TestUpdateInPastClearsDeadline(t *testing.T) {
 	}
 }
 
-func TestUpdateWithinSkewAccepted(t *testing.T) {
-	r := &fakeRecorder{}
-	w := newWatcher(50*time.Millisecond, r)
-	defer w.Close()
-
-	// 5 seconds in the past is within the 30s Skew tolerance — accept it.
-	d := time.Now().Add(-5 * time.Second)
-	if err := w.Update(d); err != nil {
-		t.Fatalf("within-skew Update should succeed, got %v", err)
-	}
-	if !w.Deadline().Equal(d) {
-		t.Fatalf("expected deadline to be applied, got %v want %v", w.Deadline(), d)
-	}
-}
-
 func TestCloseSilencesUpdates(t *testing.T) {
 	r := &fakeRecorder{}
 	w := newWatcher(50*time.Millisecond, r)
 	w.Close()
 
-	_ = w.Update(time.Now().Add(time.Hour))
-
-	time.Sleep(20 * time.Millisecond)
+	if err := w.Update(time.Now().Add(time.Hour)); err != nil {
+		t.Fatalf("Update after Close: want nil, got %v", err)
+	}
 	if got := r.snapshot(); len(got) != 0 {
 		t.Fatalf("expected no events after Close, got %+v", got)
 	}
 }
 
-// TestCloseClearsRecorderDeadline pins the profile-switch fix: a watcher
-// holding a live deadline must zero the recorder on Close so the next
-// engine's watcher (and the UI reading the shared server-scoped recorder)
-// doesn't start out showing the previous session's stale "expires in".
-func TestCloseClearsRecorderDeadline(t *testing.T) {
+// TestCloseKeepsRecorderDeadline pins the reconnect-flap fix: the watcher
+// closes on every engine restart (network change, sleep/wake) while the
+// SSO deadline stays valid across those, so Close must leave the
+// server-scoped recorder's value in place. The client run loop clears the
+// recorder when it exits for real.
+func TestCloseKeepsRecorderDeadline(t *testing.T) {
 	r := &fakeRecorder{}
 	w := newWatcher(time.Hour, r)
 
@@ -377,8 +387,8 @@ func TestCloseClearsRecorderDeadline(t *testing.T) {
 
 	w.Close()
 
-	if got := r.deadline(); !got.IsZero() {
-		t.Fatalf("recorder deadline after Close = %v, want zero", got)
+	if got := r.deadline(); !got.Equal(d) {
+		t.Fatalf("recorder deadline after Close = %v, want %v", got, d)
 	}
 }
 

client/internal/conn_mgr.go

@@ -34,6 +34,8 @@ const (
 // - Handling connection establishment based on peer signaling
 //
 // The implementation is not thread-safe; it is protected by engine.syncMsgMux.
+// 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
@@ -42,12 +44,26 @@ type ConnMgr struct {
 	rosenpassEnabled bool
 
 	lazyConnMgr *manager.Manager
+	// lazyConnMgrMu guards the lazyConnMgr pointer for readers outside the
+	// engine loop (ActivatePeer). Writers hold it in addition to
+	// engine.syncMsgMux; all other reads stay under engine.syncMsgMux only.
+	lazyConnMgrMu sync.RWMutex
+
+	// reconcileRoutedIPs re-applies a peer's routed allowed IPs after its lazy wake endpoint is
+	// (re)armed (Mode A at arm time). Injected by the engine; nil disables the reconcile.
+	reconcileRoutedIPs func(peerKey string) error
 
 	wg            sync.WaitGroup
 	lazyCtx       context.Context
 	lazyCtxCancel context.CancelFunc
 }
 
+// SetRoutedIPsReconciler injects the callback used to re-apply a peer's routed allowed IPs when
+// its lazy wake endpoint is (re)armed. Must be called before the lazy manager starts.
+func (e *ConnMgr) SetRoutedIPsReconciler(fn func(peerKey string) error) {
+	e.reconcileRoutedIPs = fn
+}
+
 func NewConnMgr(engineConfig *EngineConfig, statusRecorder *peer.Status, peerStore *peerstore.Store, iface lazyconn.WGIface) *ConnMgr {
 	e := &ConnMgr{
 		peerStore:        peerStore,
@@ -238,12 +254,20 @@ func (e *ConnMgr) RemovePeerConn(peerKey string) {
 	conn.Log.Infof("removed peer from lazy conn manager")
 }
 
+// ActivatePeer wakes an idle lazy connection. Unlike the rest of ConnMgr it is
+// safe for concurrent use: the lazy manager pointer is read under lazyConnMgrMu
+// and the manager itself is internally synchronized, so callers outside the
+// engine loop (DNS warm-up) do not need engine.syncMsgMux.
 func (e *ConnMgr) ActivatePeer(ctx context.Context, conn *peer.Conn) {
-	if !e.isStartedWithLazyMgr() {
+	e.lazyConnMgrMu.RLock()
+	lazyConnMgr := e.lazyConnMgr
+	started := lazyConnMgr != nil && e.lazyCtxCancel != nil
+	e.lazyConnMgrMu.RUnlock()
+	if !started {
 		return
 	}
 
-	if found := e.lazyConnMgr.ActivatePeer(conn.GetKey()); found {
+	if found := lazyConnMgr.ActivatePeer(conn.GetKey()); found {
 		if err := conn.Open(ctx); err != nil {
 			conn.Log.Errorf("failed to open connection: %v", err)
 		}
@@ -268,16 +292,22 @@ func (e *ConnMgr) Close() {
 
 	e.lazyCtxCancel()
 	e.wg.Wait()
+
+	e.lazyConnMgrMu.Lock()
 	e.lazyConnMgr = nil
+	e.lazyConnMgrMu.Unlock()
 }
 
 func (e *ConnMgr) initLazyManager(engineCtx context.Context) {
 	cfg := manager.Config{
 		InactivityThreshold: inactivityThresholdEnv(),
+		ReconcileAllowedIPs: e.reconcileRoutedIPs,
 	}
-	e.lazyConnMgr = manager.NewManager(cfg, engineCtx, e.peerStore, e.iface)
 
+	e.lazyConnMgrMu.Lock()
+	e.lazyConnMgr = manager.NewManager(cfg, engineCtx, e.peerStore, e.iface)
 	e.lazyCtx, e.lazyCtxCancel = context.WithCancel(engineCtx)
+	e.lazyConnMgrMu.Unlock()
 
 	e.wg.Add(1)
 	go func() {
@@ -316,7 +346,10 @@ func (e *ConnMgr) closeManager(ctx context.Context) {
 
 	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)

client/internal/conn_mgr_test.go

@@ -1,10 +1,21 @@
 package internal
 
 import (
+	"context"
+	"net"
+	"net/netip"
 	"os"
+	"sync"
 	"testing"
+	"time"
 
+	"golang.zx2c4.com/wireguard/wgctrl/wgtypes"
+
+	"github.com/netbirdio/netbird/client/iface/wgaddr"
 	"github.com/netbirdio/netbird/client/internal/lazyconn"
+	"github.com/netbirdio/netbird/client/internal/peer"
+	"github.com/netbirdio/netbird/client/internal/peerstore"
+	"github.com/netbirdio/netbird/monotime"
 )
 
 func TestResolveLazyForce(t *testing.T) {
@@ -38,3 +49,58 @@ func TestResolveLazyForce(t *testing.T) {
 		})
 	}
 }
+
+type mockLazyWGIface struct{}
+
+func (mockLazyWGIface) RemovePeer(string) error { return nil }
+func (mockLazyWGIface) UpdatePeer(string, []netip.Prefix, time.Duration, *net.UDPAddr, *wgtypes.Key) error {
+	return nil
+}
+func (mockLazyWGIface) IsUserspaceBind() bool                    { return false }
+func (mockLazyWGIface) Address() wgaddr.Address                  { return wgaddr.Address{} }
+func (mockLazyWGIface) LastActivities() map[string]monotime.Time { return nil }
+func (mockLazyWGIface) MTU() uint16                              { return 1280 }
+
+// TestConnMgr_ActivatePeerConcurrentWithLifecycle exercises ActivatePeer from
+// non-engine goroutines (the DNS warm-up path) racing the manager lifecycle,
+// which stays on the engine loop. Run with -race: it fails if ActivatePeer
+// still requires engine.syncMsgMux for safety.
+func TestConnMgr_ActivatePeerConcurrentWithLifecycle(t *testing.T) {
+	t.Setenv(lazyconn.EnvLazyConn, "on")
+
+	status := peer.NewRecorder("https://mgm")
+	store := peerstore.NewConnStore()
+	connMgr := NewConnMgr(&EngineConfig{}, status, store, mockLazyWGIface{})
+
+	conn := newTestPeerConn(t, "peerA")
+
+	ctx, cancel := context.WithCancel(context.Background())
+	defer cancel()
+	connMgr.Start(ctx)
+
+	done := make(chan struct{})
+	var wg sync.WaitGroup
+	for range 4 {
+		wg.Add(1)
+		go func() {
+			defer wg.Done()
+			for {
+				select {
+				case <-done:
+					return
+				default:
+					connMgr.ActivatePeer(ctx, conn)
+				}
+			}
+		}()
+	}
+
+	// Let the activators spin against the started manager, then tear it down
+	// underneath them and let them spin against the stopped manager.
+	time.Sleep(100 * time.Millisecond)
+	connMgr.Close()
+	time.Sleep(50 * time.Millisecond)
+
+	close(done)
+	wg.Wait()
+}

client/internal/connect.go

@@ -257,7 +257,10 @@ func (c *ConnectClient) run(mobileDependency MobileDependency, runningChan chan
 		log.Errorf("failed to clean up temporary installer file: %v", err)
 	}
 
-	defer c.statusRecorder.ClientStop()
+	defer func() {
+		c.statusRecorder.SetSessionExpiresAt(time.Time{})
+		c.statusRecorder.ClientStop()
+	}()
 	operation := func() error {
 		// if context cancelled we not start new backoff cycle
 		if c.ctx.Err() != nil {
@@ -618,6 +621,7 @@ func createEngineConfig(key wgtypes.Key, config *profilemanager.Config, peerConf
 		BlockLANAccess:      config.BlockLANAccess,
 		BlockInbound:        config.BlockInbound,
 		DisableIPv6:         config.DisableIPv6,
+		SyncMessageVersion:  config.SyncMessageVersion,
 
 		LazyConnection: lazyconn.ParseState(config.LazyConnection),
 
@@ -693,6 +697,7 @@ func loginToManagement(ctx context.Context, client mgm.Client, pubSSHKey []byte,
 		config.BlockLANAccess,
 		config.BlockInbound,
 		config.DisableIPv6,
+		config.SyncMessageVersion,
 		config.EnableSSHRoot,
 		config.EnableSSHSFTP,
 		config.EnableSSHLocalPortForwarding,

client/internal/debug/debug.go

@@ -676,6 +676,7 @@ func (g *BundleGenerator) addCommonConfigFields(configContent *strings.Builder)
 	configContent.WriteString(fmt.Sprintf("BlockLANAccess: %v\n", g.internalConfig.BlockLANAccess))
 	configContent.WriteString(fmt.Sprintf("BlockInbound: %v\n", g.internalConfig.BlockInbound))
 	configContent.WriteString(fmt.Sprintf("DisableIPv6: %v\n", g.internalConfig.DisableIPv6))
+	configContent.WriteString(fmt.Sprintf("SyncMessageVersion: %v\n", g.internalConfig.SyncMessageVersion))
 
 	if g.internalConfig.DisableNotifications != nil {
 		configContent.WriteString(fmt.Sprintf("DisableNotifications: %v\n", *g.internalConfig.DisableNotifications))

client/internal/debug/debug_test.go

@@ -887,6 +887,8 @@ func TestAddConfig_AllFieldsCovered(t *testing.T) {
 		ClientCertKeyPath:             "/tmp/key",
 		LazyConnection:                "on",
 		MTU:                           1280,
+		DisableIPv6:                   true,
+		SyncMessageVersion:            func(v int) *int { return &v }(1),
 	}
 
 	for _, anonymize := range []bool{false, true} {

client/internal/dns/local/local.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"net"
 	"net/netip"
+	"os"
 	"slices"
 	"strings"
 	"sync"
@@ -36,7 +37,43 @@ type resolver interface {
 // record is left alone (it points at something outside our mesh, e.g.
 // a non-peer upstream).
 type PeerConnectivity interface {
-	IsConnectedByIP(ip string) (known, connected bool)
+	IsConnectedByIP(ip netip.Addr) (known, connected bool)
+}
+
+// PeerActivator wakes lazy-connection peers on demand. The local resolver calls
+// it with the tunnel IPs an answer points at, so a peer that is idle (lazily
+// disconnected) starts connecting at DNS-resolution time rather than racing the
+// client's first request packet. nil disables warm-up.
+type PeerActivator interface {
+	// ActivatePeersByIP triggers wake-up for the peer(s) owning addrs and blocks
+	// until one is connected or ctx (a short per-query budget) expires. It is a
+	// fast no-op for unknown or already-connected addresses.
+	ActivatePeersByIP(ctx context.Context, addrs []netip.Addr)
+}
+
+const (
+	defaultLazyWarmupTimeout = 2 * time.Second
+	envLazyWarmupTimeout     = "NB_DNS_LAZY_WARMUP_TIMEOUT"
+)
+
+// lazyWarmupTimeoutFromEnv returns the per-query budget for waking a
+// lazy-connection peer a DNS answer points at. Tunable via
+// NB_DNS_LAZY_WARMUP_TIMEOUT (a Go duration). Parsed once at construction time.
+func lazyWarmupTimeoutFromEnv() time.Duration {
+	v := os.Getenv(envLazyWarmupTimeout)
+	if v == "" {
+		return defaultLazyWarmupTimeout
+	}
+	d, err := time.ParseDuration(v)
+	if err != nil {
+		log.Warnf("invalid %s value %q, using default %s: %v", envLazyWarmupTimeout, v, defaultLazyWarmupTimeout, err)
+		return defaultLazyWarmupTimeout
+	}
+	if d <= 0 {
+		log.Warnf("non-positive %s value %q, using default %s", envLazyWarmupTimeout, v, defaultLazyWarmupTimeout)
+		return defaultLazyWarmupTimeout
+	}
+	return d
 }
 
 type Resolver struct {
@@ -51,6 +88,12 @@ type Resolver struct {
 	// filter and preserves the legacy "return whatever is registered"
 	// behaviour for callers that never wire a status source.
 	peerConn PeerConnectivity
+	// peerActivator, when non-nil, is called at resolution time to warm the
+	// lazy connection to the peer(s) an answer points at. nil disables warm-up.
+	peerActivator PeerActivator
+	// warmupTimeout is the per-query budget for the lazy-connection warm-up
+	// wait, resolved from the environment once at construction time.
+	warmupTimeout time.Duration
 
 	ctx    context.Context
 	cancel context.CancelFunc
@@ -59,11 +102,12 @@ type Resolver struct {
 func NewResolver() *Resolver {
 	ctx, cancel := context.WithCancel(context.Background())
 	return &Resolver{
-		records: make(map[dns.Question][]dns.RR),
-		domains: make(map[domain.Domain]struct{}),
-		zones:   make(map[domain.Domain]bool),
-		ctx:     ctx,
-		cancel:  cancel,
+		records:       make(map[dns.Question][]dns.RR),
+		domains:       make(map[domain.Domain]struct{}),
+		zones:         make(map[domain.Domain]bool),
+		warmupTimeout: lazyWarmupTimeoutFromEnv(),
+		ctx:           ctx,
+		cancel:        cancel,
 	}
 }
 
@@ -76,6 +120,14 @@ func (d *Resolver) SetPeerConnectivity(p PeerConnectivity) {
 	d.peerConn = p
 }
 
+// SetPeerActivator wires the DNS-time lazy-connection warm-up. Pass nil to
+// disable. Safe to call multiple times; the latest value wins.
+func (d *Resolver) SetPeerActivator(a PeerActivator) {
+	d.mu.Lock()
+	defer d.mu.Unlock()
+	d.peerActivator = a
+}
+
 func (d *Resolver) MatchSubdomains() bool {
 	return true
 }
@@ -122,6 +174,9 @@ func (d *Resolver) ServeDNS(w dns.ResponseWriter, r *dns.Msg) {
 	replyMessage.RecursionAvailable = true
 
 	result := d.lookupRecords(logger, question)
+	// Warm before filtering: activation flips a lazily-idle target to connected,
+	// which then lets it survive the disconnected-peer filter below.
+	d.warmLazyPeers(question, result.records)
 	result.records = d.filterDisconnectedPeerAnswers(logger, question, result.records)
 	replyMessage.Authoritative = !result.hasExternalData
 	replyMessage.Answer = result.records
@@ -495,8 +550,8 @@ func (d *Resolver) filterDisconnectedPeerAnswers(logger *log.Entry, question dns
 	kept := make([]dns.RR, 0, len(records))
 	var dropped int
 	for _, rr := range records {
-		ip := extractRecordIP(rr)
-		if ip == "" {
+		ip, ok := extractRecordAddr(rr)
+		if !ok {
 			kept = append(kept, rr)
 			continue
 		}
@@ -518,22 +573,57 @@ func (d *Resolver) filterDisconnectedPeerAnswers(logger *log.Entry, question dns
 	return kept
 }
 
-// extractRecordIP returns the dotted-decimal / colon-hex IP carried by
-// an A or AAAA record, or "" for any other record type.
-func extractRecordIP(rr dns.RR) string {
+// warmLazyPeers triggers lazy-connection wake-up for the peers a resolved
+// answer points at and waits briefly for one to connect, so the caller's first
+// request doesn't race the connection establishment. Warm-up is scoped to
+// match-only (non-authoritative) zones — the synthesized private-service zones
+// and user-created zones whose records point at specific peers. The account's
+// peer zone is authoritative, so plain peer-name lookups never trigger warm-up;
+// otherwise resolving any peer's name would wake its idle connection, defeating
+// laziness mesh-wide. No-op when no activator is wired (lazy connections
+// disabled) or the answer carries no peer IPs.
+func (d *Resolver) warmLazyPeers(question dns.Question, records []dns.RR) {
+	if len(records) < 2 {
+		return
+	}
+	d.mu.RLock()
+	activator := d.peerActivator
+	var nonAuth, found bool
+	if activator != nil {
+		nonAuth, found = d.findZone(question.Name)
+	}
+	d.mu.RUnlock()
+	if activator == nil || !found || !nonAuth {
+		return
+	}
+
+	var addrs []netip.Addr
+	for _, rr := range records {
+		if addr, ok := extractRecordAddr(rr); ok {
+			addrs = append(addrs, addr)
+		}
+	}
+	if len(addrs) == 0 {
+		return
+	}
+
+	ctx, cancel := context.WithTimeout(d.ctx, d.warmupTimeout)
+	defer cancel()
+	activator.ActivatePeersByIP(ctx, addrs)
+}
+
+// extractRecordAddr returns the IP address carried by an A or AAAA record.
+// ok is false for any other record type or a record with no address.
+func extractRecordAddr(rr dns.RR) (netip.Addr, bool) {
 	switch r := rr.(type) {
 	case *dns.A:
-		if r.A == nil {
-			return ""
-		}
-		return r.A.String()
+		addr, ok := netip.AddrFromSlice(r.A)
+		return addr.Unmap(), ok
 	case *dns.AAAA:
-		if r.AAAA == nil {
-			return ""
-		}
-		return r.AAAA.String()
+		addr, ok := netip.AddrFromSlice(r.AAAA)
+		return addr.Unmap(), ok
 	}
-	return ""
+	return netip.Addr{}, false
 }
 
 // Update replaces all zones and their records

client/internal/dns/local/local_test.go

@@ -37,8 +37,8 @@ type mockPeerConnectivity struct {
 	byIP map[string]struct{ known, connected bool }
 }
 
-func (m mockPeerConnectivity) IsConnectedByIP(ip string) (known, connected bool) {
-	v, ok := m.byIP[ip]
+func (m mockPeerConnectivity) IsConnectedByIP(ip netip.Addr) (known, connected bool) {
+	v, ok := m.byIP[ip.String()]
 	if !ok {
 		return false, false
 	}

client/internal/dns/local/warmup_test.go

@@ -0,0 +1,204 @@
+package local
+
+import (
+	"context"
+	"net"
+	"net/netip"
+	"sync"
+	"testing"
+	"time"
+
+	"github.com/miekg/dns"
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/internal/dns/test"
+	nbdns "github.com/netbirdio/netbird/dns"
+)
+
+// recordingActivator records the addresses it was asked to warm and returns
+// immediately, so ServeDNS is not blocked by the test.
+type recordingActivator struct {
+	mu     sync.Mutex
+	called bool
+	addrs  []netip.Addr
+}
+
+func (r *recordingActivator) ActivatePeersByIP(_ context.Context, addrs []netip.Addr) {
+	r.mu.Lock()
+	defer r.mu.Unlock()
+	r.called = true
+	r.addrs = append(r.addrs, addrs...)
+}
+
+func serveA(t *testing.T, resolver *Resolver, name string) *dns.Msg {
+	t.Helper()
+	var resp *dns.Msg
+	w := &test.MockResponseWriter{WriteMsgFunc: func(m *dns.Msg) error { resp = m; return nil }}
+	resolver.ServeDNS(w, new(dns.Msg).SetQuestion(name, dns.TypeA))
+	return resp
+}
+
+// serviceZone registers rec in a match-only (non-authoritative) zone, the shape
+// the synthesized private-service zones arrive in.
+func serviceZone(t *testing.T, resolver *Resolver, zone string, records ...nbdns.SimpleRecord) {
+	t.Helper()
+	resolver.Update([]nbdns.CustomZone{{
+		Domain:           zone,
+		Records:          records,
+		NonAuthoritative: true,
+	}})
+}
+
+func TestLocalResolver_WarmsLazyPeerOnResolve(t *testing.T) {
+	// Warm-up fires only for multi-record answers (the HA / round-robin shape of
+	// the synthesized private-service zones), so register two peer targets.
+	const name = "svc.proxy.netbird.cloud."
+	recs := []nbdns.SimpleRecord{
+		{Name: name, Type: 1, Class: nbdns.DefaultClass, TTL: 300, RData: "100.64.0.7"},
+		{Name: name, Type: 1, Class: nbdns.DefaultClass, TTL: 300, RData: "100.64.0.8"},
+	}
+	resolver := NewResolver()
+	serviceZone(t, resolver, "proxy.netbird.cloud", recs...)
+
+	act := &recordingActivator{}
+	resolver.SetPeerActivator(act)
+
+	resp := serveA(t, resolver, name)
+	require.NotNil(t, resp, "resolver must answer")
+	require.NotEmpty(t, resp.Answer, "answer must carry the A records")
+
+	act.mu.Lock()
+	defer act.mu.Unlock()
+	assert.True(t, act.called, "activator must be invoked for a multi-record service-zone answer")
+	assert.Contains(t, act.addrs, netip.MustParseAddr("100.64.0.7"), "activator must receive the first peer IP")
+	assert.Contains(t, act.addrs, netip.MustParseAddr("100.64.0.8"), "activator must receive the second peer IP")
+}
+
+func TestLocalResolver_NoWarmupForSingleRecord(t *testing.T) {
+	// A single-record answer does not trigger warm-up; the resolver only warms
+	// multi-record answers.
+	rec := nbdns.SimpleRecord{Name: "svc.proxy.netbird.cloud.", Type: 1, Class: nbdns.DefaultClass, TTL: 300, RData: "100.64.0.7"}
+	resolver := NewResolver()
+	serviceZone(t, resolver, "proxy.netbird.cloud", rec)
+
+	act := &recordingActivator{}
+	resolver.SetPeerActivator(act)
+
+	resp := serveA(t, resolver, rec.Name)
+	require.NotNil(t, resp, "resolver must answer")
+	require.NotEmpty(t, resp.Answer, "answer must carry the A record")
+
+	act.mu.Lock()
+	defer act.mu.Unlock()
+	assert.False(t, act.called, "activator must not be invoked for a single-record answer")
+}
+
+func TestLocalResolver_NoActivatorNoWarmup(t *testing.T) {
+	// With no activator wired the resolver behaves exactly as before.
+	rec := nbdns.SimpleRecord{Name: "svc.proxy.netbird.cloud.", Type: 1, Class: nbdns.DefaultClass, TTL: 300, RData: "100.64.0.7"}
+	resolver := NewResolver()
+	serviceZone(t, resolver, "proxy.netbird.cloud", rec)
+
+	resp := serveA(t, resolver, rec.Name)
+	require.NotNil(t, resp, "resolver must still answer without an activator")
+	require.NotEmpty(t, resp.Answer, "answer must carry the A record")
+}
+
+func TestLocalResolver_NoWarmupForMissingRecord(t *testing.T) {
+	// A query that resolves to nothing must not invoke the activator (no IPs).
+	resolver := NewResolver()
+	serviceZone(t, resolver, "proxy.netbird.cloud",
+		nbdns.SimpleRecord{Name: "svc.proxy.netbird.cloud.", Type: 1, Class: nbdns.DefaultClass, TTL: 300, RData: "100.64.0.7"})
+
+	act := &recordingActivator{}
+	resolver.SetPeerActivator(act)
+
+	serveA(t, resolver, "absent.proxy.netbird.cloud.")
+
+	act.mu.Lock()
+	defer act.mu.Unlock()
+	assert.False(t, act.called, "activator must not be invoked when there is no answer")
+}
+
+func TestLocalResolver_NoWarmupInAuthoritativeZone(t *testing.T) {
+	// The account's peer zone is authoritative; resolving a peer's name there
+	// must not wake its lazy connection — warm-up is scoped to match-only
+	// (non-authoritative) zones such as the synthesized private-service zones.
+	// Use a multi-record answer so the authoritative-zone scoping is the only
+	// reason warm-up is skipped, not the single-record guard.
+	const name = "peer.netbird.cloud."
+	recs := []nbdns.SimpleRecord{
+		{Name: name, Type: 1, Class: nbdns.DefaultClass, TTL: 300, RData: "100.64.0.9"},
+		{Name: name, Type: 1, Class: nbdns.DefaultClass, TTL: 300, RData: "100.64.0.10"},
+	}
+	resolver := NewResolver()
+	resolver.Update([]nbdns.CustomZone{{
+		Domain:  "netbird.cloud",
+		Records: recs,
+	}})
+
+	act := &recordingActivator{}
+	resolver.SetPeerActivator(act)
+
+	resp := serveA(t, resolver, name)
+	require.NotNil(t, resp, "resolver must answer")
+	require.NotEmpty(t, resp.Answer, "answer must carry the A records")
+
+	act.mu.Lock()
+	defer act.mu.Unlock()
+	assert.False(t, act.called, "activator must not be invoked for authoritative-zone answers")
+}
+
+func TestLazyWarmupTimeoutFromEnv(t *testing.T) {
+	tests := []struct {
+		name   string
+		value  string
+		envSet bool
+		want   time.Duration
+	}{
+		{name: "unset uses default", want: defaultLazyWarmupTimeout},
+		{name: "valid overrides", value: "5s", envSet: true, want: 5 * time.Second},
+		{name: "invalid falls back", value: "not-a-duration", envSet: true, want: defaultLazyWarmupTimeout},
+		{name: "negative falls back", value: "-1s", envSet: true, want: defaultLazyWarmupTimeout},
+		{name: "zero falls back", value: "0s", envSet: true, want: defaultLazyWarmupTimeout},
+	}
+
+	for _, tt := range tests {
+		t.Run(tt.name, func(t *testing.T) {
+			if tt.envSet {
+				t.Setenv(envLazyWarmupTimeout, tt.value)
+			}
+			assert.Equal(t, tt.want, lazyWarmupTimeoutFromEnv())
+			assert.Equal(t, tt.want, NewResolver().warmupTimeout, "constructor must resolve the timeout once")
+		})
+	}
+}
+
+func TestExtractRecordAddr(t *testing.T) {
+	t.Run("A record yields unmapped v4", func(t *testing.T) {
+		// net.ParseIP returns the 16-byte v4-in-v6 form, the same shape
+		// miekg/dns stores after parsing an A record; the extracted address
+		// must compare equal to a plain v4 netip.Addr.
+		addr, ok := extractRecordAddr(&dns.A{A: net.ParseIP("100.64.0.7")})
+		require.True(t, ok)
+		assert.True(t, addr.Is4())
+		assert.Equal(t, netip.MustParseAddr("100.64.0.7"), addr)
+	})
+
+	t.Run("AAAA record yields v6", func(t *testing.T) {
+		addr, ok := extractRecordAddr(&dns.AAAA{AAAA: net.ParseIP("fd00::1")})
+		require.True(t, ok)
+		assert.Equal(t, netip.MustParseAddr("fd00::1"), addr)
+	})
+
+	t.Run("A record without address", func(t *testing.T) {
+		_, ok := extractRecordAddr(&dns.A{})
+		assert.False(t, ok)
+	})
+
+	t.Run("non-address record", func(t *testing.T) {
+		_, ok := extractRecordAddr(&dns.CNAME{Target: "target.netbird.cloud."})
+		assert.False(t, ok)
+	})
+}

client/internal/dns/mock_server.go

@@ -8,6 +8,7 @@ import (
 	"github.com/miekg/dns"
 
 	dnsconfig "github.com/netbirdio/netbird/client/internal/dns/config"
+	"github.com/netbirdio/netbird/client/internal/dns/local"
 	nbdns "github.com/netbirdio/netbird/dns"
 	"github.com/netbirdio/netbird/route"
 	"github.com/netbirdio/netbird/shared/management/domain"
@@ -92,6 +93,11 @@ func (m *MockServer) SetFirewall(Firewall) {
 	// Mock implementation - no-op
 }
 
+// SetPeerActivator mock implementation of SetPeerActivator from Server interface
+func (m *MockServer) SetPeerActivator(local.PeerActivator) {
+	// Mock implementation - no-op
+}
+
 // BeginBatch mock implementation of BeginBatch from Server interface
 func (m *MockServer) BeginBatch() {
 	// Mock implementation - no-op

client/internal/dns/server.go

@@ -82,6 +82,7 @@ type Server interface {
 	PopulateManagementDomain(mgmtURL *url.URL) error
 	SetRouteSources(selected, active func() route.HAMap)
 	SetFirewall(Firewall)
+	SetPeerActivator(local.PeerActivator)
 }
 
 type nsGroupsByDomain struct {
@@ -491,6 +492,13 @@ func (s *DefaultServer) SetFirewall(fw Firewall) {
 	}
 }
 
+// SetPeerActivator wires the DNS-time lazy-connection warm-up on the local
+// resolver. Injected after the connection manager exists (it does not at
+// DNS-server construction time). Pass nil to disable.
+func (s *DefaultServer) SetPeerActivator(a local.PeerActivator) {
+	s.localResolver.SetPeerActivator(a)
+}
+
 // Stop stops the server
 func (s *DefaultServer) Stop() {
 	s.ctxCancel()
@@ -1435,11 +1443,11 @@ type localPeerConnectivity struct {
 
 // IsConnectedByIP looks the IP up in the peerstore and surfaces both
 // the known and connected bits. Used by Resolver.filterDisconnectedPeerAnswers.
-func (l localPeerConnectivity) IsConnectedByIP(ip string) (known, connected bool) {
+func (l localPeerConnectivity) IsConnectedByIP(ip netip.Addr) (known, connected bool) {
 	if l.status == nil {
 		return false, false
 	}
-	state, ok := l.status.PeerStateByIP(ip)
+	state, ok := l.status.PeerStateByIP(ip.String())
 	if !ok {
 		return false, false
 	}

client/internal/dns_peer_activator.go

@@ -0,0 +1,76 @@
+package internal
+
+import (
+	"context"
+	"net/netip"
+	"time"
+
+	"github.com/netbirdio/netbird/client/internal/peer"
+	"github.com/netbirdio/netbird/client/internal/peerstore"
+)
+
+const dnsActivationPollInterval = 50 * time.Millisecond
+
+// dnsPeerActivator wakes lazy-connection peers from the DNS resolution path. It
+// implements dns/local.PeerActivator. DNS queries run on their own goroutines,
+// so it only touches state that is safe for concurrent use — ConnMgr.ActivatePeer,
+// peerstore.Store and peer.Status — and never takes the engine's syncMsgMux,
+// keeping DNS resolution from contending with network-map processing.
+type dnsPeerActivator struct {
+	connMgr   *ConnMgr
+	peerStore *peerstore.Store
+	status    *peer.Status
+	// ctx is the engine's long-lived context. The connection dial is tied to it
+	// (not the per-query DNS wait budget) so a handshake that outlasts the wait
+	// still completes in the background rather than being cancelled at the deadline.
+	ctx context.Context
+}
+
+// ActivatePeersByIP triggers wake-up for the peer(s) owning addrs and waits
+// until one is connected or ctx (the per-query DNS wait budget) expires.
+// Activation itself is tied to the engine's long-lived context so the dial
+// survives a wait that times out. Unknown or already-connected addresses are
+// skipped, so the steady-state (warm) path adds no latency.
+func (a *dnsPeerActivator) ActivatePeersByIP(ctx context.Context, addrs []netip.Addr) {
+	if a == nil || a.connMgr == nil {
+		return
+	}
+
+	var pending []string
+	for _, addr := range addrs {
+		ip := addr.String()
+		st, ok := a.status.PeerStateByIP(ip)
+		if !ok || st.ConnStatus == peer.StatusConnected {
+			continue
+		}
+		conn, ok := a.peerStore.PeerConn(st.PubKey)
+		if !ok {
+			continue
+		}
+		a.connMgr.ActivatePeer(a.ctx, conn)
+		pending = append(pending, ip)
+	}
+
+	if len(pending) == 0 {
+		return
+	}
+	a.waitConnected(ctx, pending)
+}
+
+// waitConnected blocks until any of ips reports a connected peer or ctx expires.
+func (a *dnsPeerActivator) waitConnected(ctx context.Context, ips []string) {
+	ticker := time.NewTicker(dnsActivationPollInterval)
+	defer ticker.Stop()
+	for {
+		for _, ip := range ips {
+			if st, ok := a.status.PeerStateByIP(ip); ok && st.ConnStatus == peer.StatusConnected {
+				return
+			}
+		}
+		select {
+		case <-ctx.Done():
+			return
+		case <-ticker.C:
+		}
+	}
+}

client/internal/dns_peer_activator_test.go

@@ -0,0 +1,129 @@
+package internal
+
+import (
+	"context"
+	"net/netip"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/internal/peer"
+	"github.com/netbirdio/netbird/client/internal/peerstore"
+)
+
+func newTestPeerConn(t *testing.T, key string) *peer.Conn {
+	t.Helper()
+	conn, err := peer.NewConn(peer.ConnConfig{
+		Key:      key,
+		LocalKey: "local",
+		WgConfig: peer.WgConfig{
+			AllowedIps: []netip.Prefix{netip.MustParsePrefix("100.64.0.1/32")},
+		},
+	}, peer.ServiceDependencies{})
+	require.NoError(t, err)
+	return conn
+}
+
+func newTestDNSPeerActivator(t *testing.T) (*dnsPeerActivator, *peer.Status, *peerstore.Store) {
+	t.Helper()
+	status := peer.NewRecorder("https://mgm")
+	store := peerstore.NewConnStore()
+	// ConnMgr without Start: the lazy manager is nil, so ActivatePeer is a
+	// no-op — these tests exercise the activator's skip/wait logic.
+	connMgr := NewConnMgr(&EngineConfig{}, status, store, nil)
+	return &dnsPeerActivator{
+		connMgr:   connMgr,
+		peerStore: store,
+		status:    status,
+		ctx:       context.Background(),
+	}, status, store
+}
+
+func TestDNSPeerActivator_NilSafe(t *testing.T) {
+	var a *dnsPeerActivator
+	a.ActivatePeersByIP(context.Background(), []netip.Addr{netip.MustParseAddr("100.64.0.1")})
+}
+
+// TestDNSPeerActivator_SkipsUnknownAndConnectedPeers verifies the steady-state
+// (warm) path adds no latency: already-connected and unknown addresses never
+// enter the wait loop.
+func TestDNSPeerActivator_SkipsUnknownAndConnectedPeers(t *testing.T) {
+	a, status, store := newTestDNSPeerActivator(t)
+
+	require.NoError(t, status.AddPeer("peerA", "a.netbird.cloud", "100.64.0.1", "fd00::1"))
+	require.NoError(t, status.UpdatePeerState(peer.State{PubKey: "peerA", ConnStatus: peer.StatusConnected}))
+	store.AddPeerConn("peerA", newTestPeerConn(t, "peerA"))
+
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	start := time.Now()
+	a.ActivatePeersByIP(ctx, []netip.Addr{
+		netip.MustParseAddr("100.64.0.1"),  // known, connected -> skipped
+		netip.MustParseAddr("fd00::1"),     // known via IPv6, connected -> skipped
+		netip.MustParseAddr("100.64.0.99"), // unknown -> skipped
+	})
+	require.Less(t, time.Since(start), time.Second, "no pending peer must mean no wait")
+}
+
+// TestDNSPeerActivator_WaitsForPendingPeerToConnect verifies the wait loop
+// returns as soon as a pending peer reports connected, well before the
+// per-query budget expires.
+func TestDNSPeerActivator_WaitsForPendingPeerToConnect(t *testing.T) {
+	a, status, store := newTestDNSPeerActivator(t)
+
+	require.NoError(t, status.AddPeer("peerA", "a.netbird.cloud", "100.64.0.1", ""))
+	store.AddPeerConn("peerA", newTestPeerConn(t, "peerA"))
+
+	go func() {
+		time.Sleep(150 * time.Millisecond)
+		_ = status.UpdatePeerState(peer.State{PubKey: "peerA", ConnStatus: peer.StatusConnected})
+	}()
+
+	ctx, cancel := context.WithTimeout(context.Background(), 10*time.Second)
+	defer cancel()
+
+	start := time.Now()
+	a.ActivatePeersByIP(ctx, []netip.Addr{netip.MustParseAddr("100.64.0.1")})
+	elapsed := time.Since(start)
+
+	require.GreaterOrEqual(t, elapsed, 100*time.Millisecond, "must wait for the pending peer")
+	require.Less(t, elapsed, 5*time.Second, "must return on connect, not at the deadline")
+}
+
+// TestDNSPeerActivator_ReturnsAtBudgetWhenPeerStaysIdle verifies a peer that
+// never connects releases the DNS response at the per-query budget instead of
+// blocking it indefinitely.
+func TestDNSPeerActivator_ReturnsAtBudgetWhenPeerStaysIdle(t *testing.T) {
+	a, status, store := newTestDNSPeerActivator(t)
+
+	require.NoError(t, status.AddPeer("peerA", "a.netbird.cloud", "100.64.0.1", ""))
+	store.AddPeerConn("peerA", newTestPeerConn(t, "peerA"))
+
+	ctx, cancel := context.WithTimeout(context.Background(), 300*time.Millisecond)
+	defer cancel()
+
+	start := time.Now()
+	a.ActivatePeersByIP(ctx, []netip.Addr{netip.MustParseAddr("100.64.0.1")})
+	elapsed := time.Since(start)
+
+	require.GreaterOrEqual(t, elapsed, 250*time.Millisecond, "must wait out the budget for a pending peer")
+	require.Less(t, elapsed, 5*time.Second, "must not block past the budget")
+}
+
+// TestDNSPeerActivator_NoWaitWithoutPeerConn verifies a known-but-idle peer
+// with no connection object in the store is not waited on: there is nothing to
+// activate, so waiting could only ever time out.
+func TestDNSPeerActivator_NoWaitWithoutPeerConn(t *testing.T) {
+	a, status, _ := newTestDNSPeerActivator(t)
+
+	require.NoError(t, status.AddPeer("peerA", "a.netbird.cloud", "100.64.0.1", ""))
+
+	ctx, cancel := context.WithTimeout(context.Background(), 5*time.Second)
+	defer cancel()
+
+	start := time.Now()
+	a.ActivatePeersByIP(ctx, []netip.Addr{netip.MustParseAddr("100.64.0.1")})
+	require.Less(t, time.Since(start), time.Second, "peer without a conn must not be waited on")
+}

client/internal/ebpf/ebpf/src/dns_fwd.c

@@ -52,11 +52,14 @@ int xdp_dns_fwd(struct iphdr  *ip, struct udphdr *udp) {
 
     if (udp->dest == GENERAL_DNS_PORT && ip->daddr == dns_ip) {
         udp->dest = dns_port;
+        // Clear the now-stale checksum; zero means "not computed" for IPv4.
+        udp->check = 0;
         return XDP_PASS;
     }
 
     if (udp->source == dns_port && ip->saddr == dns_ip) {
         udp->source = GENERAL_DNS_PORT;
+        udp->check = 0;
         return XDP_PASS;
     }
 

client/internal/ebpf/ebpf/src/wg_proxy.c

@@ -50,5 +50,11 @@ int xdp_wg_proxy(struct iphdr  *ip, struct udphdr *udp) {
     __be16 new_dst_port = htons(proxy_port);
     udp->dest = new_dst_port;
     udp->source = new_src_port;
+
+    // The ports are covered by the UDP checksum. This is an IPv4 loopback hop
+    // and the payload is already integrity-protected, so clear the checksum (a
+    // zero UDP checksum means "not computed" for IPv4) rather than leave a
+    // stale value the kernel would drop as UDP_CSUM.
+    udp->check = 0;
 	return XDP_PASS;
 }

client/internal/engine.go

@@ -64,7 +64,10 @@ import (
 	"github.com/netbirdio/netbird/route"
 	mgm "github.com/netbirdio/netbird/shared/management/client"
 	"github.com/netbirdio/netbird/shared/management/domain"
+	sharedgrpc "github.com/netbirdio/netbird/shared/management/grpc"
+	nbnetworkmap "github.com/netbirdio/netbird/shared/management/networkmap"
 	mgmProto "github.com/netbirdio/netbird/shared/management/proto"
+	types "github.com/netbirdio/netbird/shared/management/types"
 	"github.com/netbirdio/netbird/shared/netiputil"
 	auth "github.com/netbirdio/netbird/shared/relay/auth/hmac"
 	relayClient "github.com/netbirdio/netbird/shared/relay/client"
@@ -147,6 +150,7 @@ type EngineConfig struct {
 	BlockLANAccess      bool
 	BlockInbound        bool
 	DisableIPv6         bool
+	SyncMessageVersion  *int
 
 	// LazyConnection is the MDM-sourced lazy-connection override; StateUnset defers to
 	// the env var and management feature flag.
@@ -220,6 +224,13 @@ type Engine struct {
 	// networkSerial is the latest CurrentSerial (state ID) of the network sent by the Management service
 	networkSerial uint64
 
+	// latestComponents is the most-recent NetworkMapComponents decoded from
+	// a NetworkMapEnvelope (capability=3 peers only). Held alongside the
+	// NetworkMap that Calculate() produced from it so future incremental
+	// updates have a base to apply changes against. nil for legacy-format
+	// peers. Guarded by syncMsgMux.
+	latestComponents *types.NetworkMapComponents
+
 	networkMonitor *networkmonitor.NetworkMonitor
 
 	sshServer sshServer
@@ -652,8 +663,24 @@ func (e *Engine) Start(netbirdConfig *mgmProto.NetbirdConfig, mgmtURL *url.URL)
 	iceCfg := e.createICEConfig()
 
 	e.connMgr = NewConnMgr(e.config, e.statusRecorder, e.peerStore, wgIface)
+	e.connMgr.SetRoutedIPsReconciler(func(peerKey string) error {
+		if e.routeManager == nil {
+			return nil
+		}
+		return e.routeManager.ReconcilePeerAllowedIPs(peerKey)
+	})
 	e.connMgr.Start(e.ctx)
 
+	// Wire DNS-time lazy-connection warm-up now that the connection manager
+	// exists (it does not at DNS-server construction time). A DNS answer that
+	// points at an idle peer then wakes it before the client's first request.
+	e.dnsServer.SetPeerActivator(&dnsPeerActivator{
+		connMgr:   e.connMgr,
+		peerStore: e.peerStore,
+		status:    e.statusRecorder,
+		ctx:       e.ctx,
+	})
+
 	e.srWatcher = guard.NewSRWatcher(e.signal, e.relayManager, e.mobileDep.IFaceDiscover, iceCfg)
 	e.srWatcher.Start(peer.IsForceRelayed())
 
@@ -963,8 +990,12 @@ func (e *Engine) handleSync(update *mgmProto.SyncResponse) error {
 
 	e.ApplySessionDeadline(update.GetSessionExpiresAt())
 
-	if update.NetworkMap != nil && update.NetworkMap.PeerConfig != nil {
-		e.handleAutoUpdateVersion(update.NetworkMap.PeerConfig.AutoUpdate)
+	// Envelope sync responses carry PeerConfig at the top level; legacy
+	// NetworkMap syncs carry it under NetworkMap.PeerConfig.
+	if pc := update.GetPeerConfig(); pc != nil {
+		e.handleAutoUpdateVersion(pc.GetAutoUpdate())
+	} else if nm := update.GetNetworkMap(); nm != nil && nm.GetPeerConfig() != nil {
+		e.handleAutoUpdateVersion(nm.GetPeerConfig().GetAutoUpdate())
 	}
 
 	done := e.phase("netbird_config")
@@ -974,12 +1005,47 @@ func (e *Engine) handleSync(update *mgmProto.SyncResponse) error {
 		return err
 	}
 
+	// Decode the network map from either the components envelope or the
+	// legacy proto.NetworkMap before the posture-check gating below, so the
+	// "is there a network map" decision covers both wire shapes.
+	var (
+		nm         *mgmProto.NetworkMap
+		components *types.NetworkMapComponents
+	)
+	if version := update.GetVersion(); version == int32(sharedgrpc.ComponentNetworkMap) {
+		// Components-format peer: decode the envelope back to typed
+		// components, run Calculate() locally, and convert to the wire
+		// NetworkMap shape the rest of the engine consumes. Components are
+		// retained so future incremental updates can apply deltas instead
+		// of doing a full reconstruction.
+		envelope := update.GetNetworkMapEnvelope()
+		if envelope == nil {
+			return fmt.Errorf("received a SyncReponse indicating use of components network map, but components are missing")
+		}
+
+		localKey := e.config.WgPrivateKey.PublicKey().String()
+		dnsName := ""
+		if pc := update.GetPeerConfig(); pc != nil {
+			// PeerConfig.Fqdn = "<dns_label>.<dns_domain>" — extract the
+			// shared domain by stripping the peer's own label prefix. Falls
+			// back to empty if the FQDN doesn't have the expected shape.
+			dnsName = extractDNSDomainFromFQDN(pc.GetFqdn())
+		}
+		result, err := nbnetworkmap.EnvelopeToNetworkMap(e.ctx, envelope, localKey, dnsName)
+		if err != nil {
+			return fmt.Errorf("decode network map envelope: %w", err)
+		}
+		nm = result.NetworkMap
+		components = result.Components
+	} else {
+		nm = update.GetNetworkMap()
+	}
+
 	// Posture checks are bound to the network map presence:
 	//   NetworkMap != nil, checks present -> apply the received checks
 	//   NetworkMap != nil, checks nil     -> posture checks were removed, clear them
 	//   NetworkMap == nil                 -> config-only update (e.g. relay token rotation),
 	//                                        leave the previously applied checks untouched
-	nm := update.GetNetworkMap()
 	if nm == nil {
 		return nil
 	}
@@ -992,6 +1058,14 @@ func (e *Engine) handleSync(update *mgmProto.SyncResponse) error {
 	}
 
 	done = e.phase("persist")
+	// Only retain the components view when the server sent the envelope
+	// path. A legacy proto.NetworkMap means components == nil; writing it
+	// here would clobber a previously-cached snapshot, breaking the
+	// incremental-delta base on a future envelope sync.
+	if components != nil {
+		e.latestComponents = components
+	}
+
 	e.persistSyncResponse(update)
 	done()
 
@@ -1005,6 +1079,19 @@ func (e *Engine) handleSync(update *mgmProto.SyncResponse) error {
 	return nil
 }
 
+// extractDNSDomainFromFQDN returns the trailing dotted domain part of the
+// receiving peer's FQDN — the same value the management server fills as
+// dnsName when it builds the legacy NetworkMap. "peer42.netbird.cloud" →
+// "netbird.cloud". An empty string is returned for unrecognized formats.
+func extractDNSDomainFromFQDN(fqdn string) string {
+	for i := 0; i < len(fqdn); i++ {
+		if fqdn[i] == '.' && i+1 < len(fqdn) {
+			return fqdn[i+1:]
+		}
+	}
+	return ""
+}
+
 // updateNetbirdConfig applies the management-provided NetBird configuration:
 // STUN/TURN and relay servers, flow logging and DNS settings. A nil config is a no-op,
 // which is the case for sync updates carrying only a network map.
@@ -1164,6 +1251,7 @@ func (e *Engine) applyInfoFlags(info *system.Info) {
 		e.config.BlockLANAccess,
 		e.config.BlockInbound,
 		e.config.DisableIPv6,
+		e.config.SyncMessageVersion,
 		e.config.EnableSSHRoot,
 		e.config.EnableSSHSFTP,
 		e.config.EnableSSHLocalPortForwarding,
@@ -2032,6 +2120,7 @@ func (e *Engine) readInitialSettings() ([]*route.Route, *nbdns.Config, bool, err
 		e.config.BlockLANAccess,
 		e.config.BlockInbound,
 		e.config.DisableIPv6,
+		e.config.SyncMessageVersion,
 		e.config.EnableSSHRoot,
 		e.config.EnableSSHSFTP,
 		e.config.EnableSSHLocalPortForwarding,

client/internal/engine_session_deadline_test.go

@@ -75,4 +75,14 @@ func TestApplySessionDeadline_ThreeState(t *testing.T) {
 		require.True(t, e.statusRecorder.GetSessionExpiresAt().IsZero(),
 			"invalid timestamp must clear the deadline")
 	})
+
+	t.Run("recently expired timestamp stays visible as expired", func(t *testing.T) {
+		e := newEngine()
+		expired := time.Now().Add(-5 * time.Minute).UTC().Truncate(time.Second)
+
+		e.ApplySessionDeadline(timestamppb.New(expired))
+
+		require.True(t, e.statusRecorder.GetSessionExpiresAt().Equal(expired),
+			"recently-expired deadline must stay on the recorder so consumers render it as expired")
+	})
 }

client/internal/lazyconn/manager/manager.go

@@ -29,6 +29,11 @@ type managedPeer struct {
 
 type Config struct {
 	InactivityThreshold *time.Duration
+	// ReconcileAllowedIPs re-applies a peer's routed allowed IPs after its wake endpoint is
+	// armed. The activity listener creates the wake peer with the overlay /32 only; without the
+	// routed prefixes WireGuard would not steer subnet-bound traffic to the wake endpoint, so an
+	// idle routing peer could never be woken by that traffic. Optional; nil disables the reconcile.
+	ReconcileAllowedIPs func(peerKey string) error
 }
 
 // Manager manages lazy connections
@@ -56,6 +61,9 @@ type Manager struct {
 	peerToHAGroups map[string][]route.HAUniqueID // peer ID -> HA groups they belong to
 	haGroupToPeers map[route.HAUniqueID][]string // HA group -> peer IDs in the group
 	routesMu       sync.RWMutex
+
+	// reconcileAllowedIPs re-applies a peer's routed allowed IPs after its wake endpoint is armed.
+	reconcileAllowedIPs func(peerKey string) error
 }
 
 // NewManager creates a new lazy connection manager
@@ -73,6 +81,7 @@ func NewManager(config Config, engineCtx context.Context, peerStore *peerstore.S
 		activityManager:      activity.NewManager(wgIface),
 		peerToHAGroups:       make(map[string][]route.HAUniqueID),
 		haGroupToPeers:       make(map[route.HAUniqueID][]string),
+		reconcileAllowedIPs:  config.ReconcileAllowedIPs,
 	}
 
 	if wgIface.IsUserspaceBind() {
@@ -201,7 +210,7 @@ func (m *Manager) AddPeer(peerCfg lazyconn.PeerConfig) (bool, error) {
 		return false, nil
 	}
 
-	if err := m.activityManager.MonitorPeerActivity(peerCfg); err != nil {
+	if err := m.armActivityListener(peerCfg); err != nil {
 		return false, err
 	}
 
@@ -288,7 +297,7 @@ func (m *Manager) DeactivatePeer(peerID peerid.ConnID) {
 
 	m.inactivityManager.RemovePeer(mp.peerCfg.PublicKey)
 
-	if err := m.activityManager.MonitorPeerActivity(*mp.peerCfg); err != nil {
+	if err := m.armActivityListener(*mp.peerCfg); err != nil {
 		mp.peerCfg.Log.Errorf("failed to create activity monitor: %v", err)
 		return
 	}
@@ -465,6 +474,31 @@ func (m *Manager) close() {
 }
 
 // shouldDeferIdleForHA checks if peer should stay connected due to HA group requirements
+// armRoutedAllowedIPs re-applies the peer's routed allowed IPs onto its freshly armed wake
+// endpoint. The activity listener creates the wake peer with the overlay /32 only, so without
+// this the routed prefixes would be missing and traffic to a routed subnet could not wake the
+// idle routing peer. It is a no-op when no reconciler is configured.
+// armActivityListener (re)arms the peer's wake endpoint via the activity manager and then
+// re-applies its routed allowed IPs, so traffic to a routed subnet can wake an idle routing
+// peer. The routed prefixes must be re-applied after the wake endpoint exists because the
+// listener creates it with the overlay /32 only.
+func (m *Manager) armActivityListener(peerCfg lazyconn.PeerConfig) error {
+	if err := m.activityManager.MonitorPeerActivity(peerCfg); err != nil {
+		return err
+	}
+	m.armRoutedAllowedIPs(&peerCfg)
+	return nil
+}
+
+func (m *Manager) armRoutedAllowedIPs(peerCfg *lazyconn.PeerConfig) {
+	if m.reconcileAllowedIPs == nil {
+		return
+	}
+	if err := m.reconcileAllowedIPs(peerCfg.PublicKey); err != nil {
+		peerCfg.Log.Errorf("failed to reconcile routed allowed IPs on wake endpoint: %v", err)
+	}
+}
+
 func (m *Manager) shouldDeferIdleForHA(inactivePeers map[string]struct{}, peerID string) bool {
 	m.routesMu.RLock()
 	defer m.routesMu.RUnlock()
@@ -577,7 +611,7 @@ func (m *Manager) onPeerInactivityTimedOut(peerIDs map[string]struct{}) {
 
 		mp.peerCfg.Log.Infof("start activity monitor")
 
-		if err := m.activityManager.MonitorPeerActivity(*mp.peerCfg); err != nil {
+		if err := m.armActivityListener(*mp.peerCfg); err != nil {
 			mp.peerCfg.Log.Errorf("failed to create activity monitor: %v", err)
 			continue
 		}

client/internal/peer/status.go

@@ -813,19 +813,14 @@ func (d *Status) SetSessionExpiresAt(deadline time.Time) {
 }
 
 // GetSessionExpiresAt returns the most recently recorded SSO session deadline,
-// or the zero value when no deadline is tracked. A deadline that has already
-// slipped into the past reports as "none": once the session has expired it is
-// no longer a meaningful countdown, and the sessionwatch.Watcher does not
-// arm a timer at the deadline itself to clear it (only the two pre-expiry
-// warnings). Without this guard the UI would keep painting a stale
-// "expires in …" against a moment that has passed until the next login,
-// extend, or teardown rewrote the value.
+// or the zero value when no deadline is tracked. A deadline in the past is
+// returned as-is: it means the session has expired, and consumers (tray row,
+// CLI status) render it as "expired" rather than hiding it — masking it as
+// "none" would blank the UI at the exact moment it should say the session
+// ended.
 func (d *Status) GetSessionExpiresAt() time.Time {
 	d.mux.Lock()
 	defer d.mux.Unlock()
-	if !d.sessionExpiresAt.IsZero() && d.sessionExpiresAt.Before(time.Now()) {
-		return time.Time{}
-	}
 	return d.sessionExpiresAt
 }
 

client/internal/profilemanager/config.go

@@ -96,6 +96,7 @@ type ConfigInput struct {
 	BlockLANAccess      *bool
 	BlockInbound        *bool
 	DisableIPv6         *bool
+	SyncMessageVersion  *int
 
 	DisableNotifications *bool
 
@@ -137,6 +138,7 @@ type Config struct {
 	BlockLANAccess      bool
 	BlockInbound        bool
 	DisableIPv6         bool
+	SyncMessageVersion  *int
 
 	DisableNotifications *bool
 
@@ -587,6 +589,12 @@ func (config *Config) apply(input ConfigInput) (updated bool, err error) {
 		updated = true
 	}
 
+	if input.SyncMessageVersion != nil && *input.SyncMessageVersion != *config.SyncMessageVersion {
+		log.Infof("setting SyncMessageVersion to %v", *input.SyncMessageVersion)
+		*config.SyncMessageVersion = *input.SyncMessageVersion
+		updated = true
+	}
+
 	if input.DisableNotifications != nil && (config.DisableNotifications == nil || *input.DisableNotifications != *config.DisableNotifications) {
 		if *input.DisableNotifications {
 			log.Infof("disabling notifications")

client/internal/routemanager/dynamic/route.go

@@ -185,7 +185,7 @@ func (r *Route) startResolver(ctx context.Context) {
 }
 
 func (r *Route) update(ctx context.Context) error {
-	resolved, err := r.resolveDomains()
+	resolved, err := r.resolveDomains(ctx)
 	if err != nil {
 		if len(resolved) == 0 {
 			return fmt.Errorf("resolve domains: %w", err)
@@ -199,9 +199,9 @@ func (r *Route) update(ctx context.Context) error {
 	return nil
 }
 
-func (r *Route) resolveDomains() (domainMap, error) {
+func (r *Route) resolveDomains(ctx context.Context) (domainMap, error) {
 	results := make(chan resolveResult)
-	go r.resolve(results)
+	go r.resolve(ctx, results)
 
 	resolved := domainMap{}
 	var merr *multierror.Error
@@ -217,7 +217,7 @@ func (r *Route) resolveDomains() (domainMap, error) {
 	return resolved, nberrors.FormatErrorOrNil(merr)
 }
 
-func (r *Route) resolve(results chan resolveResult) {
+func (r *Route) resolve(ctx context.Context, results chan resolveResult) {
 	var wg sync.WaitGroup
 
 	for _, d := range r.route.Domains {
@@ -225,10 +225,10 @@ func (r *Route) resolve(results chan resolveResult) {
 		go func(domain domain.Domain) {
 			defer wg.Done()
 
-			ips, err := r.getIPsFromResolver(domain)
+			ips, err := r.getIPsFromResolver(ctx, domain)
 			if err != nil {
 				log.Tracef("Failed to resolve domain %s with private resolver: %v", domain.SafeString(), err)
-				ips, err = net.LookupIP(domain.PunycodeString())
+				ips, err = lookupHostIPs(ctx, domain)
 				if err != nil {
 					results <- resolveResult{domain: domain, err: fmt.Errorf("resolve d %s: %w", domain.SafeString(), err)}
 					return
@@ -364,6 +364,20 @@ func determinePrefixChanges(oldPrefixes, newPrefixes []netip.Prefix) (toAdd, toR
 	return
 }
 
+// lookupHostIPs resolves d via the system resolver, honoring ctx cancellation.
+func lookupHostIPs(ctx context.Context, d domain.Domain) ([]net.IP, error) {
+	addrs, err := net.DefaultResolver.LookupIPAddr(ctx, d.PunycodeString())
+	if err != nil {
+		return nil, err
+	}
+
+	ips := make([]net.IP, 0, len(addrs))
+	for _, addr := range addrs {
+		ips = append(ips, addr.IP)
+	}
+	return ips, nil
+}
+
 func combinePrefixes(oldPrefixes, removedPrefixes, addedPrefixes []netip.Prefix) []netip.Prefix {
 	prefixSet := make(map[netip.Prefix]struct{})
 	for _, prefix := range oldPrefixes {

client/internal/routemanager/dynamic/route_generic.go

@@ -3,11 +3,12 @@
 package dynamic
 
 import (
+	"context"
 	"net"
 
 	"github.com/netbirdio/netbird/shared/management/domain"
 )
 
-func (r *Route) getIPsFromResolver(domain domain.Domain) ([]net.IP, error) {
-	return net.LookupIP(domain.PunycodeString())
+func (r *Route) getIPsFromResolver(ctx context.Context, domain domain.Domain) ([]net.IP, error) {
+	return lookupHostIPs(ctx, domain)
 }

client/internal/routemanager/dynamic/route_ios.go

@@ -3,6 +3,7 @@
 package dynamic
 
 import (
+	"context"
 	"fmt"
 	"net"
 	"time"
@@ -16,7 +17,7 @@ import (
 
 const dialTimeout = 10 * time.Second
 
-func (r *Route) getIPsFromResolver(domain domain.Domain) ([]net.IP, error) {
+func (r *Route) getIPsFromResolver(ctx context.Context, domain domain.Domain) ([]net.IP, error) {
 	privateClient, err := nbdns.GetClientPrivate(r.wgInterface, r.resolverAddr.Addr(), dialTimeout)
 	if err != nil {
 		return nil, fmt.Errorf("error while creating private client: %s", err)
@@ -32,7 +33,7 @@ func (r *Route) getIPsFromResolver(domain domain.Domain) ([]net.IP, error) {
 		msg := new(dns.Msg)
 		msg.SetQuestion(fqdn, qtype)
 
-		response, _, err := nbdns.ExchangeWithFallback(nil, privateClient, msg, r.resolverAddr.String())
+		response, _, err := nbdns.ExchangeWithFallback(ctx, privateClient, msg, r.resolverAddr.String())
 		if err != nil {
 			if queryErr == nil {
 				queryErr = fmt.Errorf("DNS query for %s (type %d) after %s: %w", domain.SafeString(), qtype, time.Since(startTime), err)

client/internal/routemanager/manager.go

@@ -52,6 +52,10 @@ type Manager interface {
 	UpdateRoutes(updateSerial uint64, serverRoutes map[route.ID]*route.Route, clientRoutes route.HAMap, useNewDNSRoute bool) error
 	ClassifyRoutes(newRoutes []*route.Route) (map[route.ID]*route.Route, route.HAMap)
 	TriggerSelection(route.HAMap)
+	SelectRoutes(ids []route.NetID, appendRoute bool) error
+	DeselectRoutes(ids []route.NetID) error
+	SelectAllRoutes()
+	DeselectAllRoutes()
 	GetRouteSelector() *routeselector.RouteSelector
 	GetClientRoutes() route.HAMap
 	GetSelectedClientRoutes() route.HAMap
@@ -61,6 +65,7 @@ type Manager interface {
 	InitialRouteRange() []string
 	SetFirewall(firewall.Manager) error
 	SetDNSForwarderPort(port uint16)
+	ReconcilePeerAllowedIPs(peerKey string) error
 	Stop(stateManager *statemanager.Manager)
 }
 
@@ -232,6 +237,30 @@ func (m *DefaultManager) setupRefCounters(useNoop bool) {
 	)
 }
 
+// ReconcilePeerAllowedIPs re-applies every routed allowed IP currently tracked for the peer
+// onto the WireGuard device. The allowed-IP refcounter only calls its AddFunc (which pushes to
+// the device) on a prefix's 0->1 transition, so a peer whose device entry was rebuilt without a
+// matching refcounter change — e.g. a lazy connection cycling through idle->wake, which recreates
+// the WireGuard peer with the overlay /32 only — ends up missing routed prefixes the refcounter
+// still considers installed, and nothing retries. Calling this when the peer's WireGuard entry is
+// (re)created restores convergence. It is add-only and idempotent: AddAllowedIP is update-only, so
+// prefixes are re-added to an existing peer and an absent peer is left untouched.
+func (m *DefaultManager) ReconcilePeerAllowedIPs(peerKey string) error {
+	if m.allowedIPsRefCounter == nil {
+		return nil
+	}
+
+	return m.allowedIPsRefCounter.ReapplyMatching(
+		func(out string) bool { return out == peerKey },
+		func(prefix netip.Prefix) error {
+			if err := m.wgInterface.AddAllowedIP(peerKey, prefix); err != nil {
+				return fmt.Errorf("add allowed IP %s for peer %s: %w", prefix, peerKey, err)
+			}
+			return nil
+		},
+	)
+}
+
 // Init sets up the routing
 func (m *DefaultManager) Init() error {
 	m.routeSelector = m.initSelector()
@@ -775,7 +804,7 @@ func (m *DefaultManager) collectExitNodeInfo(clientRoutes route.HAMap) exitNodeI
 	var info exitNodeInfo
 
 	for haID, routes := range clientRoutes {
-		if !m.isExitNodeRoute(routes) {
+		if !isExitNodeRoutes(routes) {
 			continue
 		}
 
@@ -795,13 +824,6 @@ func (m *DefaultManager) collectExitNodeInfo(clientRoutes route.HAMap) exitNodeI
 	return info
 }
 
-func (m *DefaultManager) isExitNodeRoute(routes []*route.Route) bool {
-	if len(routes) == 0 {
-		return false
-	}
-	return route.IsV4DefaultRoute(routes[0].Network) || route.IsV6DefaultRoute(routes[0].Network)
-}
-
 func (m *DefaultManager) categorizeUserSelection(netID route.NetID, info *exitNodeInfo) {
 	if m.routeSelector.IsSelected(netID) {
 		info.userSelected = append(info.userSelected, netID)

client/internal/routemanager/mock.go

@@ -16,6 +16,8 @@ type MockManager struct {
 	ClassifyRoutesFunc           func(routes []*route.Route) (map[route.ID]*route.Route, route.HAMap)
 	UpdateRoutesFunc             func(updateSerial uint64, serverRoutes map[route.ID]*route.Route, clientRoutes route.HAMap, useNewDNSRoute bool) error
 	TriggerSelectionFunc         func(haMap route.HAMap)
+	SelectRoutesFunc             func(ids []route.NetID, appendRoute bool) error
+	DeselectRoutesFunc           func(ids []route.NetID) error
 	GetRouteSelectorFunc         func() *routeselector.RouteSelector
 	GetClientRoutesFunc          func() route.HAMap
 	GetSelectedClientRoutesFunc  func() route.HAMap
@@ -55,6 +57,30 @@ func (m *MockManager) TriggerSelection(networks route.HAMap) {
 	}
 }
 
+// SelectRoutes mock implementation of SelectRoutes from Manager interface
+func (m *MockManager) SelectRoutes(ids []route.NetID, appendRoute bool) error {
+	if m.SelectRoutesFunc != nil {
+		return m.SelectRoutesFunc(ids, appendRoute)
+	}
+	return nil
+}
+
+// DeselectRoutes mock implementation of DeselectRoutes from Manager interface
+func (m *MockManager) DeselectRoutes(ids []route.NetID) error {
+	if m.DeselectRoutesFunc != nil {
+		return m.DeselectRoutesFunc(ids)
+	}
+	return nil
+}
+
+// SelectAllRoutes mock implementation of SelectAllRoutes from Manager interface
+func (m *MockManager) SelectAllRoutes() {
+}
+
+// DeselectAllRoutes mock implementation of DeselectAllRoutes from Manager interface
+func (m *MockManager) DeselectAllRoutes() {
+}
+
 // GetRouteSelector mock implementation of GetRouteSelector from Manager interface
 func (m *MockManager) GetRouteSelector() *routeselector.RouteSelector {
 	if m.GetRouteSelectorFunc != nil {
@@ -112,6 +138,11 @@ func (m *MockManager) SetFirewall(firewall.Manager) error {
 func (m *MockManager) SetDNSForwarderPort(port uint16) {
 }
 
+// ReconcilePeerAllowedIPs mock implementation of ReconcilePeerAllowedIPs from Manager interface
+func (m *MockManager) ReconcilePeerAllowedIPs(peerKey string) error {
+	return nil
+}
+
 // Stop mock implementation of Stop from Manager interface
 func (m *MockManager) Stop(stateManager *statemanager.Manager) {
 	if m.StopFunc != nil {

client/internal/routemanager/reconcile_test.go

@@ -0,0 +1,90 @@
+//go:build !windows
+
+package routemanager
+
+import (
+	"net"
+	"net/netip"
+	"sync"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+	"golang.zx2c4.com/wireguard/tun/netstack"
+
+	"github.com/netbirdio/netbird/client/iface/device"
+	"github.com/netbirdio/netbird/client/iface/wgaddr"
+	"github.com/netbirdio/netbird/client/internal/routemanager/refcounter"
+)
+
+// reconcileWGMock is a minimal iface.WGIface that only records AddAllowedIP calls; every other
+// method is an inert stub because ReconcilePeerAllowedIPs exercises none of them.
+type reconcileWGMock struct {
+	mu   sync.Mutex
+	adds map[string][]netip.Prefix
+}
+
+func (m *reconcileWGMock) AddAllowedIP(peerKey string, allowedIP netip.Prefix) error {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	if m.adds == nil {
+		m.adds = map[string][]netip.Prefix{}
+	}
+	m.adds[peerKey] = append(m.adds[peerKey], allowedIP)
+	return nil
+}
+
+func (m *reconcileWGMock) added(peerKey string) []netip.Prefix {
+	m.mu.Lock()
+	defer m.mu.Unlock()
+	return m.adds[peerKey]
+}
+
+func (m *reconcileWGMock) RemoveAllowedIP(string, netip.Prefix) error { return nil }
+func (m *reconcileWGMock) Name() string                               { return "utun-test" }
+func (m *reconcileWGMock) Address() wgaddr.Address                    { return wgaddr.Address{} }
+func (m *reconcileWGMock) ToInterface() *net.Interface                { return nil }
+func (m *reconcileWGMock) IsUserspaceBind() bool                      { return false }
+func (m *reconcileWGMock) GetFilter() device.PacketFilter             { return nil }
+func (m *reconcileWGMock) GetDevice() *device.FilteredDevice          { return nil }
+func (m *reconcileWGMock) GetNet() *netstack.Net                      { return nil }
+
+// TestReconcilePeerAllowedIPs verifies the declarative reconcile re-applies every routed prefix
+// tracked for the peer (self-heal, independent of refcount level) and stays scoped to that peer.
+func TestReconcilePeerAllowedIPs(t *testing.T) {
+	wg := &reconcileWGMock{}
+	m := &DefaultManager{wgInterface: wg}
+	m.allowedIPsRefCounter = refcounter.New[netip.Prefix, string, string](
+		func(_ netip.Prefix, peerKey string) (string, error) { return peerKey, nil },
+		func(netip.Prefix, string) error { return nil },
+	)
+
+	peerA1 := netip.MustParsePrefix("10.0.0.0/24")
+	peerA2 := netip.MustParsePrefix("10.1.0.0/24")
+	peerB1 := netip.MustParsePrefix("10.2.0.0/24")
+
+	for prefix, peer := range map[netip.Prefix]string{peerA1: "peerA", peerA2: "peerA", peerB1: "peerB"} {
+		_, err := m.allowedIPsRefCounter.Increment(prefix, peer)
+		require.NoError(t, err)
+	}
+	// Extra reference: reconcile must still re-apply the prefix even though its refcount never
+	// hit 0 again (the exact case the plain incremental path skips).
+	_, err := m.allowedIPsRefCounter.Increment(peerA1, "peerA")
+	require.NoError(t, err)
+
+	require.NoError(t, m.ReconcilePeerAllowedIPs("peerA"))
+
+	assert.ElementsMatch(t, []netip.Prefix{peerA1, peerA2}, wg.added("peerA"),
+		"reconcile must re-apply all routed prefixes of the peer")
+	assert.Empty(t, wg.added("peerB"), "reconcile must not touch another peer's prefixes")
+}
+
+// TestReconcilePeerAllowedIPsNoCounter verifies reconcile is a safe no-op before the refcounter is
+// set up.
+func TestReconcilePeerAllowedIPsNoCounter(t *testing.T) {
+	wg := &reconcileWGMock{}
+	m := &DefaultManager{wgInterface: wg}
+
+	require.NoError(t, m.ReconcilePeerAllowedIPs("peerA"))
+	assert.Empty(t, wg.added("peerA"))
+}

client/internal/routemanager/refcounter/refcounter.go

@@ -94,6 +94,26 @@ func (rm *Counter[Key, I, O]) Get(key Key) (Ref[O], bool) {
 	return ref, ok
 }
 
+// ReapplyMatching calls apply for every key whose stored Out satisfies pred, holding the
+// counter lock for the whole pass. Running apply under the lock keeps it atomic with respect
+// to Increment/Decrement: a prefix dropped to zero is removed from the map (and had its
+// RemoveFunc called) before this pass observes it, so a stale key can never be re-applied.
+// pred and apply are invoked under the lock, so they must not call back into the counter.
+func (rm *Counter[Key, I, O]) ReapplyMatching(pred func(out O) bool, apply func(key Key) error) error {
+	rm.mu.Lock()
+	defer rm.mu.Unlock()
+
+	var merr *multierror.Error
+	for key, ref := range rm.refCountMap {
+		if pred(ref.Out) {
+			if err := apply(key); err != nil {
+				merr = multierror.Append(merr, err)
+			}
+		}
+	}
+	return nberrors.FormatErrorOrNil(merr)
+}
+
 // Increment increments the reference count for the given key.
 // If this is the first reference to the key, the AddFunc is called.
 func (rm *Counter[Key, I, O]) Increment(key Key, in I) (Ref[O], error) {

client/internal/routemanager/refcounter/refcounter_test.go

@@ -0,0 +1,47 @@
+package refcounter
+
+import (
+	"net/netip"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+)
+
+// TestReapplyMatching verifies ReapplyMatching invokes apply for exactly the keys whose stored
+// Out satisfies the predicate (no duplicates for multiply-referenced keys) — the primitive
+// ReconcilePeerAllowedIPs relies on to re-apply a single peer's routed prefixes.
+func TestReapplyMatching(t *testing.T) {
+	rc := New[netip.Prefix, string, string](
+		func(_ netip.Prefix, peerKey string) (string, error) { return peerKey, nil },
+		func(netip.Prefix, string) error { return nil },
+	)
+
+	peerA1 := netip.MustParsePrefix("10.0.0.0/24")
+	peerA2 := netip.MustParsePrefix("10.1.0.0/24")
+	peerB1 := netip.MustParsePrefix("10.2.0.0/24")
+
+	for prefix, peer := range map[netip.Prefix]string{peerA1: "peerA", peerA2: "peerA", peerB1: "peerB"} {
+		_, err := rc.Increment(prefix, peer)
+		require.NoError(t, err)
+	}
+	// a second reference must not make the key applied twice
+	_, err := rc.Increment(peerA1, "peerA")
+	require.NoError(t, err)
+
+	var applied []netip.Prefix
+	err = rc.ReapplyMatching(
+		func(out string) bool { return out == "peerA" },
+		func(key netip.Prefix) error { applied = append(applied, key); return nil },
+	)
+	require.NoError(t, err)
+	assert.ElementsMatch(t, []netip.Prefix{peerA1, peerA2}, applied)
+
+	var none []netip.Prefix
+	err = rc.ReapplyMatching(
+		func(out string) bool { return out == "missing" },
+		func(key netip.Prefix) error { none = append(none, key); return nil },
+	)
+	require.NoError(t, err)
+	assert.Empty(t, none)
+}

client/internal/routemanager/selection.go

@@ -0,0 +1,138 @@
+package routemanager
+
+import (
+	"fmt"
+	"slices"
+
+	"github.com/hashicorp/go-multierror"
+	log "github.com/sirupsen/logrus"
+	"golang.org/x/exp/maps"
+
+	nberrors "github.com/netbirdio/netbird/client/errors"
+	"github.com/netbirdio/netbird/route"
+)
+
+// SelectRoutes selects the routes with the given network IDs and applies the
+// new selection. V4/v6 exit-node pairs are expanded automatically. Exit nodes
+// are mutually exclusive: if the selection activates an exit node, every other
+// available exit node is deselected so two can't be active at once. With
+// appendRoute=false the previous selection is replaced instead of extended.
+func (m *DefaultManager) SelectRoutes(ids []route.NetID, appendRoute bool) error {
+	if err := m.selectRoutes(ids, appendRoute); err != nil {
+		return err
+	}
+	m.TriggerSelection(m.GetClientRoutes())
+	return nil
+}
+
+// DeselectRoutes removes the routes with the given network IDs from the
+// selection and applies the change. V4/v6 exit-node pairs are expanded
+// automatically.
+func (m *DefaultManager) DeselectRoutes(ids []route.NetID) error {
+	if err := m.deselectRoutes(ids); err != nil {
+		return err
+	}
+	m.TriggerSelection(m.GetClientRoutes())
+	return nil
+}
+
+func (m *DefaultManager) deselectRoutes(ids []route.NetID) error {
+	routesMap := m.GetClientRoutesWithNetID()
+	routes := route.ExpandV6ExitPairs(slices.Clone(ids), routesMap)
+
+	log.Debugf("deselecting routes with ids: %v", routes)
+
+	if err := m.routeSelector.DeselectRoutes(routes, maps.Keys(routesMap)); err != nil {
+		return fmt.Errorf("deselect routes: %w", err)
+	}
+
+	return nil
+}
+
+// SelectAllRoutes selects every available route and applies the selection.
+// Exit nodes stay mutually exclusive: at most one remains active.
+func (m *DefaultManager) SelectAllRoutes() {
+	m.selectAllRoutes()
+	m.TriggerSelection(m.GetClientRoutes())
+}
+
+func (m *DefaultManager) selectAllRoutes() {
+	m.routeSelector.SelectAllRoutes()
+
+	// Select-all wipes every explicit selection, so exit nodes fall back to
+	// management's auto-apply flags — which may mark several at once.
+	// Reconcile immediately so at most one exit node stays active instead of
+	// waiting for the next network map to enforce it.
+	m.mux.Lock()
+	defer m.mux.Unlock()
+	m.updateRouteSelectorFromManagement(m.clientRoutes)
+}
+
+// DeselectAllRoutes deselects every route and applies the change.
+func (m *DefaultManager) DeselectAllRoutes() {
+	m.routeSelector.DeselectAllRoutes()
+	m.TriggerSelection(m.GetClientRoutes())
+}
+
+func (m *DefaultManager) selectRoutes(ids []route.NetID, appendRoute bool) error {
+	routesMap := m.GetClientRoutesWithNetID()
+	routes := route.ExpandV6ExitPairs(slices.Clone(ids), routesMap)
+	allIDs := maps.Keys(routesMap)
+
+	log.Debugf("selecting routes with ids: %v", routes)
+
+	// A partial failure (e.g. an unknown ID in the request) still selects the
+	// valid routes, so exclusivity below must run regardless of the error.
+	var merr *multierror.Error
+	if err := m.routeSelector.SelectRoutes(routes, appendRoute, allIDs); err != nil {
+		merr = multierror.Append(merr, fmt.Errorf("select routes: %w", err))
+	}
+
+	// Exit nodes are mutually exclusive: if this selection activates an
+	// exit node, deselect every other available exit node so two can't be
+	// selected at once. Non-exit route selections are left untouched.
+	if requestActivatesExitNode(routes, routesMap) {
+		if others := otherExitNodeIDs(routesMap, routes); len(others) > 0 {
+			if err := m.routeSelector.DeselectRoutes(others, allIDs); err != nil {
+				merr = multierror.Append(merr, fmt.Errorf("deselect sibling exit nodes: %w", err))
+			}
+		}
+	}
+
+	return nberrors.FormatErrorOrNil(merr)
+}
+
+func isExitNodeRoutes(routes []*route.Route) bool {
+	return len(routes) > 0 && (route.IsV4DefaultRoute(routes[0].Network) || route.IsV6DefaultRoute(routes[0].Network))
+}
+
+// requestActivatesExitNode reports whether any requested NetID maps to an exit
+// node (default route) in the current route table.
+func requestActivatesExitNode(requested []route.NetID, routesMap map[route.NetID][]*route.Route) bool {
+	for _, id := range requested {
+		if isExitNodeRoutes(routesMap[id]) {
+			return true
+		}
+	}
+	return false
+}
+
+// otherExitNodeIDs returns every available exit-node NetID that is not in the
+// requested set — the siblings to deselect so a single exit node stays active.
+func otherExitNodeIDs(routesMap map[route.NetID][]*route.Route, requested []route.NetID) []route.NetID {
+	keep := make(map[route.NetID]struct{}, len(requested))
+	for _, id := range requested {
+		keep[id] = struct{}{}
+	}
+	var others []route.NetID
+	for id, routes := range routesMap {
+		if !isExitNodeRoutes(routes) {
+			continue
+		}
+		if _, ok := keep[id]; ok {
+			continue
+		}
+		others = append(others, id)
+	}
+	return others
+}

client/internal/routemanager/selection_test.go

@@ -0,0 +1,129 @@
+package routemanager
+
+import (
+	"net/netip"
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/client/internal/routeselector"
+	"github.com/netbirdio/netbird/route"
+)
+
+func v6ExitRoute(netID, peer string) *route.Route {
+	return &route.Route{
+		NetID:   route.NetID(netID),
+		Network: netip.MustParsePrefix("::/0"),
+		Peer:    peer,
+	}
+}
+
+func newSelectionTestManager() *DefaultManager {
+	return &DefaultManager{
+		routeSelector: routeselector.NewRouteSelector(),
+		clientRoutes: route.HAMap{
+			"exitA|0.0.0.0/0":    {exitRoute("exitA", "p1", true)},
+			"exitA-v6|::/0":      {v6ExitRoute("exitA-v6", "p1")},
+			"exitB|0.0.0.0/0":    {exitRoute("exitB", "p2", true)},
+			"lan|192.168.1.0/24": {{NetID: "lan", Network: netip.MustParsePrefix("192.168.1.0/24"), Peer: "p3"}},
+		},
+	}
+}
+
+func TestSelectRoutes_ExitNodeExclusivity(t *testing.T) {
+	m := newSelectionTestManager()
+
+	// Selecting an exit node selects its v6 pair and deselects the sibling.
+	require.NoError(t, m.selectRoutes([]route.NetID{"exitA"}, true))
+	assert.True(t, m.routeSelector.IsSelected("exitA"), "exitA should be selected")
+	assert.True(t, m.routeSelector.IsSelected("exitA-v6"), "the v6 pair follows its v4 base")
+	assert.False(t, m.routeSelector.IsSelected("exitB"), "the sibling exit node must be deselected")
+
+	// Switching to the sibling deselects the previous exit node and its v6 pair.
+	require.NoError(t, m.selectRoutes([]route.NetID{"exitB"}, true))
+	assert.True(t, m.routeSelector.IsSelected("exitB"), "exitB should now be selected")
+	assert.False(t, m.routeSelector.IsSelected("exitA"), "the previous exit node must be deselected")
+	assert.False(t, m.routeSelector.IsSelected("exitA-v6"), "the previous exit node's v6 pair must be deselected")
+	assert.True(t, m.routeSelector.IsSelected("lan"), "non-exit route selection is untouched")
+
+	// Selecting a non-exit route leaves the active exit node alone.
+	require.NoError(t, m.selectRoutes([]route.NetID{"lan"}, true))
+	assert.True(t, m.routeSelector.IsSelected("exitB"), "selecting a non-exit route keeps the exit node")
+
+	// Deselecting the active exit node turns every exit node off.
+	require.NoError(t, m.deselectRoutes([]route.NetID{"exitB"}))
+	assert.False(t, m.routeSelector.IsSelected("exitB"), "exitB should be deselected")
+	assert.False(t, m.routeSelector.IsSelected("exitA"), "exitA stays deselected")
+	assert.True(t, m.routeSelector.IsSelected("lan"), "non-exit route selection is untouched")
+}
+
+func TestSelectRoutes_PartialErrorStillEnforcesExclusivity(t *testing.T) {
+	// The unknown ID must be reported, but the valid exit node in the same
+	// request is still selected — so its sibling must still be deselected.
+	// Both orderings are covered: processing must continue past the invalid
+	// ID wherever it sits in the request.
+	requests := map[string][]route.NetID{
+		"invalid id first": {"missing", "exitB"},
+		"invalid id last":  {"exitB", "missing"},
+	}
+
+	for name, ids := range requests {
+		t.Run(name, func(t *testing.T) {
+			m := newSelectionTestManager()
+
+			require.NoError(t, m.selectRoutes([]route.NetID{"exitA"}, true))
+
+			err := m.selectRoutes(ids, true)
+			assert.Error(t, err, "unknown id must be reported")
+			assert.True(t, m.routeSelector.IsSelected("exitB"), "valid exit node from the request is selected")
+			assert.False(t, m.routeSelector.IsSelected("exitA"), "sibling exit node must be deselected despite the error")
+			assert.False(t, m.routeSelector.IsSelected("exitA-v6"), "sibling's v6 pair must be deselected too")
+		})
+	}
+}
+
+func TestSelectAllRoutes_KeepsSingleExitNode(t *testing.T) {
+	// Both exit nodes are marked for auto-apply by management
+	// (SkipAutoApply=false), the state where select-all could turn on two at
+	// once without the immediate reconciliation.
+	m := &DefaultManager{
+		routeSelector: routeselector.NewRouteSelector(),
+		clientRoutes: route.HAMap{
+			"exitA|0.0.0.0/0":    {exitRoute("exitA", "p1", false)},
+			"exitB|0.0.0.0/0":    {exitRoute("exitB", "p2", false)},
+			"lan|192.168.1.0/24": {{NetID: "lan", Network: netip.MustParsePrefix("192.168.1.0/24"), Peer: "p3"}},
+		},
+	}
+
+	require.NoError(t, m.selectRoutes([]route.NetID{"exitB"}, true))
+
+	m.selectAllRoutes()
+
+	assert.True(t, m.routeSelector.IsSelected("lan"), "non-exit routes are all selected")
+	assert.True(t, m.routeSelector.IsSelected("exitA"), "the deterministic management pick stays active")
+	assert.False(t, m.routeSelector.IsSelected("exitB"), "select-all must not leave a second exit node active")
+}
+
+func TestSelectRoutes_UnknownRoute(t *testing.T) {
+	m := newSelectionTestManager()
+
+	assert.Error(t, m.selectRoutes([]route.NetID{"missing"}, true), "selecting an unavailable route must fail")
+	assert.Error(t, m.deselectRoutes([]route.NetID{"missing"}), "deselecting an unavailable route must fail")
+}
+
+func TestExitNodeSelectionHelpers(t *testing.T) {
+	routesMap := map[route.NetID][]*route.Route{
+		"exitA": {{Network: netip.MustParsePrefix("0.0.0.0/0")}},
+		"exitB": {{Network: netip.MustParsePrefix("::/0")}},
+		"lan":   {{Network: netip.MustParsePrefix("192.168.0.0/16")}},
+	}
+
+	assert.True(t, requestActivatesExitNode([]route.NetID{"exitA"}, routesMap), "v4 default route is an exit node")
+	assert.True(t, requestActivatesExitNode([]route.NetID{"exitB"}, routesMap), "v6 default route is an exit node")
+	assert.False(t, requestActivatesExitNode([]route.NetID{"lan"}, routesMap), "lan route is not an exit node")
+	assert.False(t, requestActivatesExitNode([]route.NetID{"missing"}, routesMap), "unknown id is not an exit node")
+
+	others := otherExitNodeIDs(routesMap, []route.NetID{"exitB"})
+	assert.ElementsMatch(t, []route.NetID{"exitA"}, others, "only the other exit node is a sibling; the lan route is ignored")
+}

client/ios/NetBirdSDK/client.go

@@ -13,7 +13,6 @@ import (
 	"time"
 
 	log "github.com/sirupsen/logrus"
-	"golang.org/x/exp/maps"
 
 	"github.com/netbirdio/netbird/client/internal"
 	"github.com/netbirdio/netbird/client/internal/auth"
@@ -637,23 +636,18 @@ func (c *Client) SelectRoute(id string) error {
 	}
 
 	routeManager := engine.GetRouteManager()
-	routeSelector := routeManager.GetRouteSelector()
 	if id == "All" {
 		log.Debugf("select all routes")
-		routeSelector.SelectAllRoutes()
-	} else {
-		log.Debugf("select route with id: %s", id)
-		routes := toNetIDs([]string{id})
-		routesMap := routeManager.GetClientRoutesWithNetID()
-		routes = route.ExpandV6ExitPairs(routes, routesMap)
-		if err := routeSelector.SelectRoutes(routes, true, maps.Keys(routesMap)); err != nil {
-			log.Debugf("error when selecting routes: %s", err)
-			return fmt.Errorf("select routes: %w", err)
-		}
+		routeManager.SelectAllRoutes()
+		return nil
 	}
-	routeManager.TriggerSelection(routeManager.GetClientRoutes())
-	return nil
 
+	log.Debugf("select route with id: %s", id)
+	if err := routeManager.SelectRoutes(toNetIDs([]string{id}), true); err != nil {
+		log.Debugf("error when selecting routes: %s", err)
+		return err
+	}
+	return nil
 }
 
 func (c *Client) DeselectRoute(id string) error {
@@ -667,21 +661,17 @@ func (c *Client) DeselectRoute(id string) error {
 	}
 
 	routeManager := engine.GetRouteManager()
-	routeSelector := routeManager.GetRouteSelector()
 	if id == "All" {
 		log.Debugf("deselect all routes")
-		routeSelector.DeselectAllRoutes()
-	} else {
-		log.Debugf("deselect route with id: %s", id)
-		routes := toNetIDs([]string{id})
-		routesMap := routeManager.GetClientRoutesWithNetID()
-		routes = route.ExpandV6ExitPairs(routes, routesMap)
-		if err := routeSelector.DeselectRoutes(routes, maps.Keys(routesMap)); err != nil {
-			log.Debugf("error when deselecting routes: %s", err)
-			return fmt.Errorf("deselect routes: %w", err)
-		}
+		routeManager.DeselectAllRoutes()
+		return nil
+	}
+
+	log.Debugf("deselect route with id: %s", id)
+	if err := routeManager.DeselectRoutes(toNetIDs([]string{id})); err != nil {
+		log.Debugf("error when deselecting routes: %s", err)
+		return err
 	}
-	routeManager.TriggerSelection(routeManager.GetClientRoutes())
 	return nil
 }
 

client/ios/NetBirdSDK/version.go

@@ -0,0 +1,12 @@
+//go:build ios
+
+package NetBirdSDK
+
+import "github.com/netbirdio/netbird/version"
+
+// GoClientVersion returns the NetBird Go client version that was baked into
+// the framework at compile time via
+// -ldflags "-X github.com/netbirdio/netbird/version.version=<version>".
+func GoClientVersion() string {
+	return version.NetbirdVersion()
+}

client/server/network.go

@@ -8,7 +8,6 @@ import (
 	"sort"
 	"strings"
 
-	"golang.org/x/exp/maps"
 	"google.golang.org/grpc/codes"
 	gstatus "google.golang.org/grpc/status"
 
@@ -161,30 +160,11 @@ func (s *Server) SelectNetworks(_ context.Context, req *proto.SelectNetworksRequ
 		return nil, fmt.Errorf("no route manager")
 	}
 
-	routeSelector := routeManager.GetRouteSelector()
 	if req.GetAll() {
-		routeSelector.SelectAllRoutes()
-	} else {
-		routes := toNetIDs(req.GetNetworkIDs())
-		routesMap := routeManager.GetClientRoutesWithNetID()
-		routes = route.ExpandV6ExitPairs(routes, routesMap)
-		netIdRoutes := maps.Keys(routesMap)
-		if err := routeSelector.SelectRoutes(routes, req.GetAppend(), netIdRoutes); err != nil {
-			return nil, fmt.Errorf("select routes: %w", err)
-		}
-
-		// Exit nodes are mutually exclusive: if this selection activates an
-		// exit node, deselect every other available exit node so two can't be
-		// selected at once. Non-exit route selections are left untouched.
-		if requestActivatesExitNode(routes, routesMap) {
-			if others := otherExitNodeIDs(routesMap, routes); len(others) > 0 {
-				if err := routeSelector.DeselectRoutes(others, netIdRoutes); err != nil {
-					return nil, fmt.Errorf("deselect sibling exit nodes: %w", err)
-				}
-			}
-		}
+		routeManager.SelectAllRoutes()
+	} else if err := routeManager.SelectRoutes(toNetIDs(req.GetNetworkIDs()), req.GetAppend()); err != nil {
+		return nil, err
 	}
-	routeManager.TriggerSelection(routeManager.GetClientRoutes())
 
 	s.statusRecorder.PublishEvent(
 		proto.SystemEvent_INFO,
@@ -224,19 +204,11 @@ func (s *Server) DeselectNetworks(_ context.Context, req *proto.SelectNetworksRe
 		return nil, fmt.Errorf("no route manager")
 	}
 
-	routeSelector := routeManager.GetRouteSelector()
 	if req.GetAll() {
-		routeSelector.DeselectAllRoutes()
-	} else {
-		routes := toNetIDs(req.GetNetworkIDs())
-		routesMap := routeManager.GetClientRoutesWithNetID()
-		routes = route.ExpandV6ExitPairs(routes, routesMap)
-		netIdRoutes := maps.Keys(routesMap)
-		if err := routeSelector.DeselectRoutes(routes, netIdRoutes); err != nil {
-			return nil, fmt.Errorf("deselect routes: %w", err)
-		}
+		routeManager.DeselectAllRoutes()
+	} else if err := routeManager.DeselectRoutes(toNetIDs(req.GetNetworkIDs())); err != nil {
+		return nil, err
 	}
-	routeManager.TriggerSelection(routeManager.GetClientRoutes())
 
 	s.statusRecorder.PublishEvent(
 		proto.SystemEvent_INFO,
@@ -261,37 +233,3 @@ func toNetIDs(routes []string) []route.NetID {
 	return netIDs
 }
 
-func isExitNodeRoutes(routes []*route.Route) bool {
-	return len(routes) > 0 && (route.IsV4DefaultRoute(routes[0].Network) || route.IsV6DefaultRoute(routes[0].Network))
-}
-
-// requestActivatesExitNode reports whether any requested NetID maps to an exit
-// node (default route) in the current route table.
-func requestActivatesExitNode(requested []route.NetID, routesMap map[route.NetID][]*route.Route) bool {
-	for _, id := range requested {
-		if isExitNodeRoutes(routesMap[id]) {
-			return true
-		}
-	}
-	return false
-}
-
-// otherExitNodeIDs returns every available exit-node NetID that is not in the
-// requested set — the siblings to deselect so a single exit node stays active.
-func otherExitNodeIDs(routesMap map[route.NetID][]*route.Route, requested []route.NetID) []route.NetID {
-	keep := make(map[route.NetID]struct{}, len(requested))
-	for _, id := range requested {
-		keep[id] = struct{}{}
-	}
-	var others []route.NetID
-	for id, routes := range routesMap {
-		if !isExitNodeRoutes(routes) {
-			continue
-		}
-		if _, ok := keep[id]; ok {
-			continue
-		}
-		others = append(others, id)
-	}
-	return others
-}

client/server/network_exitnode_test.go

@@ -1,26 +0,0 @@
-package server
-
-import (
-	"net/netip"
-	"testing"
-
-	"github.com/stretchr/testify/assert"
-
-	"github.com/netbirdio/netbird/route"
-)
-
-func TestExitNodeSelectionHelpers(t *testing.T) {
-	routesMap := map[route.NetID][]*route.Route{
-		"exitA": {{Network: netip.MustParsePrefix("0.0.0.0/0")}},
-		"exitB": {{Network: netip.MustParsePrefix("::/0")}},
-		"lan":   {{Network: netip.MustParsePrefix("192.168.0.0/16")}},
-	}
-
-	assert.True(t, requestActivatesExitNode([]route.NetID{"exitA"}, routesMap), "v4 default route is an exit node")
-	assert.True(t, requestActivatesExitNode([]route.NetID{"exitB"}, routesMap), "v6 default route is an exit node")
-	assert.False(t, requestActivatesExitNode([]route.NetID{"lan"}, routesMap), "lan route is not an exit node")
-	assert.False(t, requestActivatesExitNode([]route.NetID{"missing"}, routesMap), "unknown id is not an exit node")
-
-	others := otherExitNodeIDs(routesMap, []route.NetID{"exitB"})
-	assert.ElementsMatch(t, []route.NetID{"exitA"}, others, "only the other exit node is a sibling; the lan route is ignored")
-}

client/server/server.go

@@ -1081,7 +1081,10 @@ func (s *Server) Down(ctx context.Context, _ *proto.DownRequest) (*proto.DownRes
 
 	if err := s.cleanupConnection(); err != nil {
 		s.mutex.Unlock()
-		// todo review to update the status in case any type of error
+		if errors.Is(err, ErrServiceNotUp) {
+			log.Debugf("Down called while service not up: %v", err)
+			return nil, err
+		}
 		log.Errorf("failed to shut down properly: %v", err)
 		return nil, err
 	}
@@ -1154,7 +1157,7 @@ func (s *Server) cleanupConnection() error {
 	// making the run loop the sole owner of engine shutdown.
 	if engine != nil {
 		if err := engine.Stop(); err != nil {
-			return err
+			log.Errorf("failed to stop engine during cleanup: %v", err)
 		}
 	}
 

client/system/info.go

@@ -79,13 +79,15 @@ type Info struct {
 	EnableSSHLocalPortForwarding  bool
 	EnableSSHRemotePortForwarding bool
 	DisableSSHAuth                bool
+
+	SyncMessageVersion *int
 }
 
 func (i *Info) SetFlags(
 	rosenpassEnabled, rosenpassPermissive bool,
 	serverSSHAllowed *bool,
 	disableClientRoutes, disableServerRoutes,
-	disableDNS, disableFirewall, blockLANAccess, blockInbound, disableIPv6 bool,
+	disableDNS, disableFirewall, blockLANAccess, blockInbound, disableIPv6 bool, syncMessageVersion *int,
 	enableSSHRoot, enableSSHSFTP, enableSSHLocalPortForwarding, enableSSHRemotePortForwarding *bool,
 	disableSSHAuth *bool,
 ) {
@@ -103,6 +105,8 @@ func (i *Info) SetFlags(
 	i.BlockInbound = blockInbound
 	i.DisableIPv6 = disableIPv6
 
+	i.SyncMessageVersion = syncMessageVersion
+
 	if enableSSHRoot != nil {
 		i.EnableSSHRoot = *enableSSHRoot
 	}

client/ui/autostart_default.go

@@ -51,7 +51,7 @@ func autostartDisabledByMDM(policy *mdm.Policy) bool {
 // netbirdFootprintExists reports whether the machine already carries NetBird
 // daemon config or state, meaning this is not a genuinely fresh install. It is
 // the update-safety gate for the autostart default: upgrading users always
-// have a footprint, so an update can never trigger a login-item write.
+// have a footprint, so an update can never trigger a autostart entry write.
 func netbirdFootprintExists() bool {
 	candidates := []string{
 		profilemanager.DefaultConfigPath,
@@ -69,9 +69,23 @@ func netbirdFootprintExists() bool {
 // applyAutostartDefault runs the one-time launch-on-login default for genuinely
 // fresh installs. The autostartInitialized marker is persisted before any
 // enable attempt so a crash mid-flow degrades to "never enabled" instead of
-// retrying login-item writes on every launch. A user's later disable in
+// retrying autostart entry writes on every launch. A user's later disable in
 // Settings is never overridden: the marker guarantees at-most-once, ever.
 func applyAutostartDefault(ctx context.Context, autostart *services.Autostart, prefs *preferences.Store, prefsFileExisted bool) {
+	mdmDisabled := autostartDisabledByMDM(mdm.LoadPolicy())
+
+	if mdmDisabled {
+		if enabled, err := autostart.IsEnabled(ctx); err != nil {
+			log.Warnf("MDM disableAutostart: read autostart state: %v", err)
+		} else if enabled {
+			if err := autostart.SetEnabled(ctx, false); err != nil {
+				log.Warnf("MDM disableAutostart: force off failed: %v", err)
+			} else {
+				log.Info("MDM disableAutostart enforced: autostart turned off")
+			}
+		}
+	}
+
 	priorFootprint := netbirdFootprintExists() || prefsFileExisted
 
 	if prefs.Get().AutostartInitialized {
@@ -84,7 +98,7 @@ func applyAutostartDefault(ctx context.Context, autostart *services.Autostart, p
 
 	state := autostartDefaultState{
 		supported:    autostart.Supported(ctx),
-		mdmDisabled:  autostartDisabledByMDM(mdm.LoadPolicy()),
+		mdmDisabled:  mdmDisabled,
 		priorInstall: priorFootprint,
 	}
 	enable, reason := shouldEnableAutostartDefault(state)

client/ui/frontend/src/components/empty-state/DaemonOutdatedOverlay.tsx

@@ -1,10 +1,13 @@
+import { useEffect, useState } from "react";
 import { useTranslation } from "react-i18next";
 import { AlertTriangleIcon, DownloadIcon } from "lucide-react";
 import { Browser } from "@wailsio/runtime";
+import { Version } from "@bindings/services";
 import { Button } from "@/components/buttons/Button";
 import { useStatus } from "@/contexts/StatusContext.tsx";
 
 const RELEASES_URL = "https://github.com/netbirdio/netbird/releases/latest";
+const RC_RELEASES_URL = "https://pkgs.netbird.io/releases/rc";
 
 function openUrl(url: string) {
     Browser.OpenURL(url).catch(() => globalThis.open(url, "_blank"));
@@ -12,7 +15,26 @@ function openUrl(url: string) {
 
 export const DaemonOutdatedOverlay = () => {
     const { t } = useTranslation();
-    const { isDaemonOutdated } = useStatus();
+    const { status, isDaemonOutdated } = useStatus();
+
+    const [guiVersion, setGuiVersion] = useState<string>("-");
+    const clientVersion = status?.daemonVersion ?? "—";
+
+    const isRc = /-rc/i.test(guiVersion) || /-rc/i.test(clientVersion);
+    const downloadUrl = isRc ? RC_RELEASES_URL : RELEASES_URL;
+
+    useEffect(() => {
+        if (!isDaemonOutdated) return;
+        let cancelled = false;
+        Version.GUI()
+            .then((v) => {
+                if (!cancelled) setGuiVersion(v);
+            })
+            .catch((err) => console.error("[DaemonOutdatedOverlay] GUI version error", err));
+        return () => {
+            cancelled = true;
+        };
+    }, [isDaemonOutdated]);
 
     if (!isDaemonOutdated) return null;
 
@@ -38,10 +60,37 @@ export const DaemonOutdatedOverlay = () => {
                     <p className={"text-sm text-nb-gray-300"}>{t("daemon.outdated.description")}</p>
                 </div>
 
+                <div className={"flex flex-col items-center gap-0.5 text-center"}>
+                    <p className={"text-sm font-semibold text-nb-gray-100"}>
+                        {clientVersion === "development" ? (
+                            <span>
+                                {t("settings.about.clientName")}{" "}
+                                <span className={"font-mono text-yellow-400"}>
+                                    {t("settings.about.development")}
+                                </span>
+                            </span>
+                        ) : (
+                            t("settings.about.client", { version: clientVersion })
+                        )}
+                    </p>
+                    <p className={"text-sm font-medium text-nb-gray-250"}>
+                        {guiVersion === "development" ? (
+                            <span>
+                                {t("settings.about.guiName")}{" "}
+                                <span className={"font-mono text-yellow-400"}>
+                                    {t("settings.about.development")}
+                                </span>
+                            </span>
+                        ) : (
+                            t("settings.about.gui", { version: guiVersion })
+                        )}
+                    </p>
+                </div>
+
                 <div className={"wails-no-draggable"}>
-                    <Button variant={"primary"} size={"xs"} onClick={() => openUrl(RELEASES_URL)}>
+                    <Button variant={"primary"} size={"xs"} onClick={() => openUrl(downloadUrl)}>
                         <DownloadIcon size={14} />
-                        {t("update.card.getInstaller")}
+                        {t("daemon.outdated.download")}
                     </Button>
                 </div>
             </div>

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

@@ -28,6 +28,7 @@ type ProfileContextValue = {
     loaded: boolean;
     refresh: () => Promise<void>;
     switchProfile: (id: string) => Promise<void>;
+    switchProfileNoConnect: (id: string) => Promise<void>;
     addProfile: (name: string) => Promise<string>;
     removeProfile: (id: string) => Promise<void>;
     renameProfile: (id: string, newName: string) => Promise<void>;
@@ -112,6 +113,16 @@ export const ProfileProvider = ({ children }: { children: ReactNode }) => {
         [username, refresh],
     );
 
+    // Manage-profiles variant: switches without connecting, so the user can
+    // still adjust the management URL before bringing the connection up.
+    const switchProfileNoConnect = useCallback(
+        async (id: string) => {
+            await ProfileSwitcher.SwitchActiveNoConnect({ profileName: id, username });
+            await refresh();
+        },
+        [username, refresh],
+    );
+
     // addProfile creates a profile by display name and returns the
     // daemon-generated ID, so the caller can immediately address it by ID.
     const addProfile = useCallback(
@@ -158,6 +169,7 @@ export const ProfileProvider = ({ children }: { children: ReactNode }) => {
             loaded,
             refresh,
             switchProfile,
+            switchProfileNoConnect,
             addProfile,
             removeProfile,
             renameProfile,
@@ -171,6 +183,7 @@ export const ProfileProvider = ({ children }: { children: ReactNode }) => {
             loaded,
             refresh,
             switchProfile,
+            switchProfileNoConnect,
             addProfile,
             removeProfile,
             renameProfile,

client/ui/frontend/src/modules/auto-update/UpdateVersionCard.tsx

@@ -2,6 +2,7 @@ import { type ReactNode } from "react";
 import { useTranslation } from "react-i18next";
 import { Browser } from "@wailsio/runtime";
 import { DownloadIcon, NotepadText } from "lucide-react";
+import { Update as UpdateSvc } from "@bindings/services";
 import { Button } from "@/components/buttons/Button";
 import { useClientVersion } from "@/contexts/ClientVersionContext";
 import { cn } from "@/lib/cn";
@@ -14,6 +15,12 @@ function openUrl(url: string) {
     });
 }
 
+function openInstallerDownload() {
+    UpdateSvc.DownloadURL()
+        .then(openUrl)
+        .catch(() => openUrl(GITHUB_RELEASES));
+}
+
 export function UpdateVersionCard() {
     const { t } = useTranslation();
     const { updateVersion, enforced, triggerUpdate } = useClientVersion();
@@ -37,11 +44,7 @@ export function UpdateVersionCard() {
                         {t("update.card.installNow")}
                     </Button>
                 ) : (
-                    <Button
-                        variant={"primary"}
-                        size={"xs"}
-                        onClick={() => openUrl(GITHUB_RELEASES)}
-                    >
+                    <Button variant={"primary"} size={"xs"} onClick={openInstallerDownload}>
                         <DownloadIcon size={14} />
                         {t("update.card.getInstaller")}
                     </Button>

client/ui/frontend/src/modules/profiles/ProfilesTab.tsx

@@ -45,7 +45,7 @@ export function ProfilesTab() {
         activeProfileId,
         loaded,
         username,
-        switchProfile,
+        switchProfileNoConnect,
         addProfile,
         removeProfile,
         renameProfile,
@@ -100,7 +100,7 @@ export function ProfilesTab() {
             confirmLabel: t("profile.switch.confirm"),
         });
         if (!ok) return;
-        await guarded(i18next.t("profile.error.switchTitle"), () => switchProfile(id));
+        await guarded(i18next.t("profile.error.switchTitle"), () => switchProfileNoConnect(id));
     };
 
     const handleDeregister = async (id: string, name: string) => {
@@ -129,14 +129,13 @@ export function ProfilesTab() {
         await guarded(i18next.t("profile.error.createTitle"), async () => {
             const id = await addProfile(name);
             // SetConfig is keyed by the new profile's ID, so it writes the
-            // not-yet-active profile. Write before switching so any reconnect
-            // targets the right deployment.
+            // not-yet-active profile before the switch makes it current.
             if (!isNetbirdCloud(managementUrl)) {
                 await SettingsSvc.SetConfig(
                     new SetConfigParams({ profileName: id, username, managementUrl }),
                 );
             }
-            await switchProfile(id);
+            await switchProfileNoConnect(id);
         });
     };
 

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

@@ -73,6 +73,13 @@ export default function SessionExpirationDialog() {
 
         let offCancel: (() => void) | undefined;
 
+        // Return the dialog to its interactive state and dismiss the browser popup
+        const resetDialog = () => {
+            offCancel?.();
+            WindowManager.CloseBrowserLogin().catch(console.error);
+            setBusy(false);
+        };
+
         try {
             const start = await Session.RequestExtend({ hint: "" });
             const uri = start.verificationUriComplete || start.verificationUri;
@@ -105,25 +112,22 @@ export default function SessionExpirationDialog() {
             if (outcome.kind === "cancel") {
                 waitPromise.cancel?.();
                 waitPromise.catch(() => {});
+                resetDialog();
                 return;
             }
 
             // Another surface owns this flow; keep the dialog open to retry.
             if (outcome.result.preempted) {
+                resetDialog();
                 return;
             }
-
-            // Close before the popup so the restore can't flash this window back.
-            WindowManager.CloseSessionExpiration().catch(console.error);
+            WindowManager.CloseRenewFlow().catch(console.error);
         } catch (e) {
+            resetDialog();
             await errorDialog({
                 Title: t("sessionExpiration.extendFailedTitle"),
                 Message: formatErrorMessage(e),
             });
-        } finally {
-            offCancel?.();
-            WindowManager.CloseBrowserLogin().catch(console.error);
-            setBusy(false);
         }
     }, [busy, t]);
 
@@ -139,12 +143,11 @@ export default function SessionExpirationDialog() {
             });
             WindowManager.CloseSessionExpiration().catch(console.error);
         } catch (e) {
+            setBusy(false);
             await errorDialog({
                 Title: t("sessionExpiration.logoutFailedTitle"),
                 Message: formatErrorMessage(e),
             });
-        } finally {
-            setBusy(false);
         }
     }, [busy, t]);
 

client/ui/frontend/src/modules/welcome/WelcomeStepTray.tsx

@@ -22,6 +22,9 @@ type WelcomeStepTrayProps = {
 export function WelcomeStepTray({ onContinue }: Readonly<WelcomeStepTrayProps>) {
     const { t } = useTranslation();
     const trayScreenshot = trayScreenshotForOS();
+    // macOS has no tray — the icon sits in the menu bar, so the copy says so.
+    const titleKey = isMacOS() ? "welcome.titleMac" : "welcome.title";
+    const descriptionKey = isMacOS() ? "welcome.descriptionMac" : "welcome.description";
 
     return (
         <>
@@ -36,9 +39,9 @@ export function WelcomeStepTray({ onContinue }: Readonly<WelcomeStepTrayProps>)
 
             <div className={"flex w-full flex-col gap-1"}>
                 <DialogHeading id={"nb-welcome-title"} align={"left"}>
-                    {t("welcome.title")}
+                    {t(titleKey)}
                 </DialogHeading>
-                <DialogDescription align={"left"}>{t("welcome.description")}</DialogDescription>
+                <DialogDescription align={"left"}>{t(descriptionKey)}</DialogDescription>
             </div>
 
             <DialogActions>

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

@@ -1034,9 +1034,15 @@
     "welcome.title": {
         "message": "Suchen Sie NetBird in der Taskleiste"
     },
+    "welcome.titleMac": {
+        "message": "Suchen Sie NetBird in der Menüleiste"
+    },
     "welcome.description": {
         "message": "NetBird läuft in Ihrer Taskleiste. Klicken Sie auf das Symbol, um sich zu verbinden, Profile zu wechseln oder die Einstellungen zu öffnen."
     },
+    "welcome.descriptionMac": {
+        "message": "NetBird läuft in Ihrer Menüleiste. Klicken Sie auf das Symbol, um sich zu verbinden, Profile zu wechseln oder die Einstellungen zu öffnen."
+    },
     "welcome.continue": {
         "message": "Weiter"
     },
@@ -1293,10 +1299,13 @@
         "message": "Dokumentation"
     },
     "daemon.outdated.title": {
-        "message": "NetBird-Dienst ist veraltet"
+        "message": "NetBird Client ist veraltet"
     },
     "daemon.outdated.description": {
-        "message": "Aktualisieren Sie den NetBird-Dienst, um diese App zu verwenden."
+        "message": "Die neue GUI ist nicht mit Ihrem älteren Client kompatibel. Aktualisieren Sie Ihren Client, um die neue Anwendung zu verwenden."
+    },
+    "daemon.outdated.download": {
+        "message": "Neueste Version herunterladen"
     },
     "error.jwt_clock_skew": {
         "message": "Anmeldung fehlgeschlagen: Die Uhr dieses Geräts ist nicht mit dem Server synchron. Bitte synchronisieren Sie die Systemuhr und versuchen Sie es erneut."

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

@@ -1377,11 +1377,19 @@
     },
     "welcome.title": {
         "message": "Look for NetBird in your tray",
-        "description": "Heading on the first onboarding step, pointing the user to the tray icon. 'tray' = system tray / menu bar."
+        "description": "Heading on the first onboarding step, pointing the user to the tray icon. Shown on Windows and Linux; macOS uses welcome.titleMac."
+    },
+    "welcome.titleMac": {
+        "message": "Look for NetBird in your menu bar",
+        "description": "Heading on the first onboarding step on macOS, pointing the user to the menu bar icon. Use your language's Apple term for the macOS menu bar."
     },
     "welcome.description": {
         "message": "NetBird lives in your tray. Click the icon to connect, switch profiles, or open settings.",
-        "description": "Body of the first onboarding step explaining the tray icon."
+        "description": "Body of the first onboarding step explaining the tray icon. Shown on Windows and Linux; macOS uses welcome.descriptionMac."
+    },
+    "welcome.descriptionMac": {
+        "message": "NetBird lives in your menu bar. Click the icon to connect, switch profiles, or open settings.",
+        "description": "Body of the first onboarding step on macOS explaining the menu bar icon. Use your language's Apple term for the macOS menu bar."
     },
     "welcome.continue": {
         "message": "Continue",
@@ -1724,12 +1732,16 @@
         "description": "Documentation link on the daemon-unavailable overlay."
     },
     "daemon.outdated.title": {
-        "message": "NetBird Service Is Outdated",
-        "description": "Title of the overlay shown when the NetBird background service is too old to drive this UI."
+        "message": "NetBird Client Is Outdated",
+        "description": "Title of the overlay shown when the NetBird client (daemon) is too old to drive this UI."
     },
     "daemon.outdated.description": {
-        "message": "Update the NetBird service to use this app.",
-        "description": "Body of the daemon-outdated overlay telling the user to upgrade the service."
+        "message": "The new GUI isn't compatible with the older NetBird client. Update your client to use the new application.",
+        "description": "Body of the daemon-outdated overlay explaining that the GUI is newer than the client and the client must be updated."
+    },
+    "daemon.outdated.download": {
+        "message": "Download Latest",
+        "description": "Button on the daemon-outdated overlay that opens the download page for the latest release."
     },
     "error.jwt_clock_skew": {
         "message": "Sign-in failed: this device's clock is out of sync with the server. Please sync your system clock and try again.",

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

@@ -1034,9 +1034,15 @@
     "welcome.title": {
         "message": "Busque NetBird en su bandeja del sistema"
     },
+    "welcome.titleMac": {
+        "message": "Busque NetBird en su barra de menús"
+    },
     "welcome.description": {
         "message": "NetBird reside en su bandeja del sistema. Haga clic en el icono para conectarse, cambiar de perfil o abrir la configuración."
     },
+    "welcome.descriptionMac": {
+        "message": "NetBird reside en su barra de menús. Haga clic en el icono para conectarse, cambiar de perfil o abrir la configuración."
+    },
     "welcome.continue": {
         "message": "Continuar"
     },
@@ -1293,10 +1299,13 @@
         "message": "Documentación"
     },
     "daemon.outdated.title": {
-        "message": "El servicio de NetBird está desactualizado"
+        "message": "NetBird Client está desactualizado"
     },
     "daemon.outdated.description": {
-        "message": "Actualice el servicio de NetBird para usar esta aplicación."
+        "message": "La nueva GUI no es compatible con su cliente anterior. Actualice su cliente para usar la nueva aplicación."
+    },
+    "daemon.outdated.download": {
+        "message": "Descargar la última versión"
     },
     "error.jwt_clock_skew": {
         "message": "Error al iniciar sesión: el reloj de este dispositivo no está sincronizado con el servidor. Sincronice el reloj del sistema e inténtelo de nuevo."

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

@@ -1034,9 +1034,15 @@
     "welcome.title": {
         "message": "Cherchez NetBird dans votre barre d’état système"
     },
+    "welcome.titleMac": {
+        "message": "Cherchez NetBird dans votre barre des menus"
+    },
     "welcome.description": {
         "message": "NetBird se trouve dans votre barre d’état système. Cliquez sur l’icône pour vous connecter, changer de profil ou ouvrir les paramètres."
     },
+    "welcome.descriptionMac": {
+        "message": "NetBird se trouve dans votre barre des menus. Cliquez sur l’icône pour vous connecter, changer de profil ou ouvrir les paramètres."
+    },
     "welcome.continue": {
         "message": "Continuer"
     },
@@ -1293,10 +1299,13 @@
         "message": "Documentation"
     },
     "daemon.outdated.title": {
-        "message": "Le service NetBird est obsolète"
+        "message": "Le Client NetBird est obsolète"
     },
     "daemon.outdated.description": {
-        "message": "Mettez à jour le service NetBird pour utiliser cette application."
+        "message": "La nouvelle GUI n'est pas compatible avec votre ancien client. Mettez à jour votre client pour utiliser la nouvelle application."
+    },
+    "daemon.outdated.download": {
+        "message": "Télécharger la dernière version"
     },
     "error.jwt_clock_skew": {
         "message": "Échec de la connexion : l’horloge de cet appareil n’est pas synchronisée avec le serveur. Veuillez synchroniser l’horloge de votre système et réessayer."

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

@@ -1034,9 +1034,15 @@
     "welcome.title": {
         "message": "Keresse a NetBirdöt a tálcán"
     },
+    "welcome.titleMac": {
+        "message": "Keresse a NetBirdöt a menüsorban"
+    },
     "welcome.description": {
         "message": "A NetBird a tálcán fut. Kattintson az ikonra a csatlakozáshoz, profilváltáshoz vagy a beállítások megnyitásához."
     },
+    "welcome.descriptionMac": {
+        "message": "A NetBird a menüsorban fut. Kattintson az ikonra a csatlakozáshoz, profilváltáshoz vagy a beállítások megnyitásához."
+    },
     "welcome.continue": {
         "message": "Folytatás"
     },
@@ -1293,10 +1299,13 @@
         "message": "Dokumentáció"
     },
     "daemon.outdated.title": {
-        "message": "A NetBird szolgáltatás elavult"
+        "message": "A NetBird Kliens elavult"
     },
     "daemon.outdated.description": {
-        "message": "Frissítsd a NetBird szolgáltatást az alkalmazás használatához."
+        "message": "Az új GUI nem kompatibilis a régebbi klienseddel. Frissítsd a klienst az új alkalmazás használatához."
+    },
+    "daemon.outdated.download": {
+        "message": "Legújabb letöltése"
     },
     "error.jwt_clock_skew": {
         "message": "A bejelentkezés sikertelen: az eszköz órája eltér a szerverétől. Kérjük, szinkronizálja a rendszer óráját, majd próbálja újra."

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

@@ -1034,9 +1034,15 @@
     "welcome.title": {
         "message": "Cerchi NetBird nella tray"
     },
+    "welcome.titleMac": {
+        "message": "Cerchi NetBird nella barra dei menu"
+    },
     "welcome.description": {
         "message": "NetBird risiede nella tray. Clicchi sull'icona per connettersi, cambiare profilo o aprire le impostazioni."
     },
+    "welcome.descriptionMac": {
+        "message": "NetBird risiede nella barra dei menu. Clicchi sull'icona per connettersi, cambiare profilo o aprire le impostazioni."
+    },
     "welcome.continue": {
         "message": "Continua"
     },
@@ -1293,10 +1299,13 @@
         "message": "Documentazione"
     },
     "daemon.outdated.title": {
-        "message": "Il servizio NetBird è obsoleto"
+        "message": "NetBird Client è obsoleto"
     },
     "daemon.outdated.description": {
-        "message": "Aggiorna il servizio NetBird per usare questa app."
+        "message": "La nuova GUI non è compatibile con il tuo client precedente. Aggiorna il client per usare la nuova applicazione."
+    },
+    "daemon.outdated.download": {
+        "message": "Scarica l'ultima versione"
     },
     "error.jwt_clock_skew": {
         "message": "Accesso non riuscito: l'orologio di questo dispositivo non è sincronizzato con il server. Sincronizzi l'orologio di sistema e riprovi."

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

@@ -1034,9 +1034,15 @@
     "welcome.title": {
         "message": "トレイの NetBird を確認してください"
     },
+    "welcome.titleMac": {
+        "message": "メニューバーの NetBird を確認してください"
+    },
     "welcome.description": {
         "message": "NetBird はトレイに常駐します。アイコンをクリックして、接続、プロファイルの切り替え、設定を開くことができます。"
     },
+    "welcome.descriptionMac": {
+        "message": "NetBird はメニューバーに常駐します。アイコンをクリックして、接続、プロファイルの切り替え、設定を開くことができます。"
+    },
     "welcome.continue": {
         "message": "続ける"
     },

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

@@ -1034,9 +1034,15 @@
     "welcome.title": {
         "message": "Procure o NetBird na sua bandeja"
     },
+    "welcome.titleMac": {
+        "message": "Procure o NetBird na sua barra de menus"
+    },
     "welcome.description": {
         "message": "O NetBird fica na sua bandeja. Clique no ícone para conectar, alternar perfis ou abrir as configurações."
     },
+    "welcome.descriptionMac": {
+        "message": "O NetBird fica na sua barra de menus. Clique no ícone para conectar, alternar perfis ou abrir as configurações."
+    },
     "welcome.continue": {
         "message": "Continuar"
     },
@@ -1293,10 +1299,13 @@
         "message": "Documentação"
     },
     "daemon.outdated.title": {
-        "message": "O serviço NetBird está desatualizado"
+        "message": "O NetBird Client está desatualizado"
     },
     "daemon.outdated.description": {
-        "message": "Atualize o serviço NetBird para usar este aplicativo."
+        "message": "A nova GUI não é compatível com o seu cliente mais antigo. Atualize o seu cliente para usar o novo aplicativo."
+    },
+    "daemon.outdated.download": {
+        "message": "Baixar a versão mais recente"
     },
     "error.jwt_clock_skew": {
         "message": "Falha no login: o relógio deste dispositivo está fora de sincronia com o servidor. Sincronize o relógio do sistema e tente novamente."

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

@@ -1034,9 +1034,15 @@
     "welcome.title": {
         "message": "Найдите NetBird в системном трее"
     },
+    "welcome.titleMac": {
+        "message": "Найдите NetBird в строке меню"
+    },
     "welcome.description": {
         "message": "NetBird находится в системном трее. Нажмите на значок, чтобы подключиться, переключить профиль или открыть настройки."
     },
+    "welcome.descriptionMac": {
+        "message": "NetBird находится в строке меню. Нажмите на значок, чтобы подключиться, переключить профиль или открыть настройки."
+    },
     "welcome.continue": {
         "message": "Продолжить"
     },
@@ -1293,10 +1299,13 @@
         "message": "Документация"
     },
     "daemon.outdated.title": {
-        "message": "Служба NetBird устарела"
+        "message": "Клиент NetBird устарел"
     },
     "daemon.outdated.description": {
-        "message": "Обновите службу NetBird, чтобы использовать это приложение."
+        "message": "Новый GUI несовместим с вашим более старым клиентом. Обновите клиент, чтобы использовать новое приложение."
+    },
+    "daemon.outdated.download": {
+        "message": "Скачать последнюю версию"
     },
     "error.jwt_clock_skew": {
         "message": "Не удалось войти: часы этого устройства рассинхронизированы с сервером. Синхронизируйте системные часы и повторите попытку."

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

@@ -1034,9 +1034,15 @@
     "welcome.title": {
         "message": "在托盘中查找 NetBird"
     },
+    "welcome.titleMac": {
+        "message": "在菜单栏中查找 NetBird"
+    },
     "welcome.description": {
         "message": "NetBird 驻留在您的托盘中。点击图标即可连接、切换配置文件或打开设置。"
     },
+    "welcome.descriptionMac": {
+        "message": "NetBird 驻留在您的菜单栏中。点击图标即可连接、切换配置文件或打开设置。"
+    },
     "welcome.continue": {
         "message": "继续"
     },
@@ -1293,10 +1299,13 @@
         "message": "文档"
     },
     "daemon.outdated.title": {
-        "message": "NetBird 服务版本过旧"
+        "message": "NetBird 客户端版本过旧"
     },
     "daemon.outdated.description": {
-        "message": "请更新 NetBird 服务以使用此应用。"
+        "message": "新版 GUI 与您较旧的客户端不兼容。请更新客户端以使用新应用。"
+    },
+    "daemon.outdated.download": {
+        "message": "下载最新版本"
     },
     "error.jwt_clock_skew": {
         "message": "登录失败：此设备的时钟与服务器不同步。请同步您的系统时钟后重试。"

client/ui/services/autostart.go

@@ -10,7 +10,7 @@ import (
 	"github.com/wailsapp/wails/v3/pkg/application"
 )
 
-// Autostart facade over Wails' AutostartManager. The OS login-item registration
+// Autostart facade over Wails' AutostartManager. The OS autostart entry registration
 // is the single source of truth; nothing is mirrored to preferences.
 type Autostart struct {
 	mgr *application.AutostartManager

client/ui/services/profileswitcher.go

@@ -12,13 +12,15 @@ import (
 	"github.com/netbirdio/netbird/client/internal/profilemanager"
 )
 
-// ProfileSwitcher holds the reconnect policy shared by the tray and React
-// frontend so both flip profiles identically. The policy keys off prevStatus
-// from DaemonFeed.Get at SwitchActive entry:
+// ProfileSwitcher holds the switch policy shared by the tray and React
+// frontend so both flip profiles identically. SwitchActive (plain selection:
+// header dropdown, tray submenu) always connects after the switch;
+// SwitchActiveNoConnect (manage-profiles screen) never does, so the user can
+// still adjust the management URL before connecting. prevStatus from
+// DaemonFeed.Get at entry only decides the teardown:
 //
-//	Connected/Connecting → Switch + Down + Up; optimistic Connecting paint.
-//	NeedsLogin/LoginFailed/SessionExpired → Switch + Down; clear stale error for re-login.
-//	Idle → Switch only.
+//	Connected/Connecting/NeedsLogin/LoginFailed/SessionExpired → Down first.
+//	Idle → no Down.
 type ProfileSwitcher struct {
 	profiles   *Profiles
 	connection *Connection
@@ -29,29 +31,40 @@ func NewProfileSwitcher(profiles *Profiles, connection *Connection, feed *Daemon
 	return &ProfileSwitcher{profiles: profiles, connection: connection, feed: feed}
 }
 
-// SwitchActive switches to the named profile applying the reconnect policy.
+// SwitchActive switches to the named profile and always connects afterwards.
 func (s *ProfileSwitcher) SwitchActive(ctx context.Context, p ProfileRef) error {
+	return s.switchActive(ctx, p, true)
+}
+
+// SwitchActiveNoConnect switches to the named profile without connecting,
+// tearing down any existing connection first.
+func (s *ProfileSwitcher) SwitchActiveNoConnect(ctx context.Context, p ProfileRef) error {
+	return s.switchActive(ctx, p, false)
+}
+
+func (s *ProfileSwitcher) switchActive(ctx context.Context, p ProfileRef, connect bool) error {
 	prevStatus := ""
-	if st, err := s.feed.Get(ctx); err == nil {
-		prevStatus = st.Status
-	} else {
-		log.Warnf("profileswitcher: get status: %v", err)
+	if s.feed != nil {
+		if st, err := s.feed.Get(ctx); err == nil {
+			prevStatus = st.Status
+		} else {
+			log.Warnf("profileswitcher: get status: %v", err)
+		}
 	}
 
-	wasActive := strings.EqualFold(prevStatus, StatusConnected) ||
-		strings.EqualFold(prevStatus, StatusConnecting)
-	needsDown := wasActive ||
+	needsDown := strings.EqualFold(prevStatus, StatusConnected) ||
+		strings.EqualFold(prevStatus, StatusConnecting) ||
 		strings.EqualFold(prevStatus, StatusNeedsLogin) ||
 		strings.EqualFold(prevStatus, StatusLoginFailed) ||
 		strings.EqualFold(prevStatus, StatusSessionExpired)
 
-	log.Infof("profileswitcher: switch profile=%q prevStatus=%q wasActive=%v needsDown=%v",
-		p.ProfileName, prevStatus, wasActive, needsDown)
+	log.Infof("profileswitcher: switch profile=%q prevStatus=%q connect=%v needsDown=%v",
+		p.ProfileName, prevStatus, connect, needsDown)
 
-	// Optimistic Connecting paint only when wasActive: those prevStatuses emit
-	// stale Connected + transient Idle pushes during Down that must be
-	// suppressed until Up resumes the stream (see DaemonFeed suppression table).
-	if wasActive {
+	// Optimistic Connecting paint plus stale-push suppression during Down (see
+	// DaemonFeed suppression table); also arms the login-watch that pops
+	// browser-login when the new profile turns out to need SSO.
+	if connect && s.feed != nil {
 		s.feed.BeginProfileSwitch()
 	}
 
@@ -76,9 +89,9 @@ func (s *ProfileSwitcher) SwitchActive(ctx context.Context, p ProfileRef) error
 		}
 	}
 
-	if wasActive {
+	if connect {
 		if err := s.connection.Up(ctx, UpParams(p)); err != nil {
-			return fmt.Errorf("reconnect %q: %w", p.ProfileName, err)
+			return fmt.Errorf("connect %q: %w", p.ProfileName, err)
 		}
 	}
 

client/ui/services/update.go

@@ -10,6 +10,7 @@ import (
 
 	"github.com/netbirdio/netbird/client/proto"
 	"github.com/netbirdio/netbird/client/ui/updater"
+	"github.com/netbirdio/netbird/version"
 )
 
 // UpdateResult mirrors TriggerUpdateResponse.
@@ -33,6 +34,12 @@ func (s *Update) GetState() updater.State {
 	return s.holder.Get()
 }
 
+// DownloadURL returns the platform-appropriate installer download link for
+// manual (non-enforced) updates.
+func (s *Update) DownloadURL() string {
+	return version.DownloadUrl()
+}
+
 // Quit exits the app. Scheduled off the calling goroutine so the JS caller's
 // response returns before the runtime tears down.
 func (s *Update) Quit() {

client/ui/services/windowmanager.go

@@ -185,37 +185,38 @@ func (s *WindowManager) OpenBrowserLogin(uri string) {
 			startURL = "/#/dialog/browser-login?uri=" + url.QueryEscape(uri)
 		}
 		s.hideOtherWindowsLocked("browser-login")
-		// Prefer the main window's screen (multi-monitor); falls back to OS-default centering.
-		var screen *application.Screen
-		if s.mainWindow != nil {
-			if sc, err := s.mainWindow.GetScreen(); err == nil {
-				screen = sc
-			}
-		}
 		opts := DialogWindowOptions("browser-login", s.title("window.title.signIn"), startURL, s.linuxIcon)
 		// Not always-on-top: it would obscure the browser tab the user logs in through.
 		opts.AlwaysOnTop = false
 		opts.InitialPosition = application.WindowCentered
-		opts.Screen = screen
+		// Open on the active (where users cursor is) display, like the session-expiration dialog.
+		opts.Screen = s.getScreenBasedOnCursorPosition()
 		s.browserLogin = s.app.Window.NewWithOptions(opts)
 		bl := s.browserLogin
-		// Red-X close means cancel: emit the event so startLogin() tears down the SSO wait.
 		bl.OnWindowEvent(events.Common.WindowClosing, func(_ *application.WindowEvent) {
-			s.app.Event.Emit(EventBrowserLoginCancel)
 			s.mu.Lock()
-			s.browserLogin = nil
-			s.restoreHiddenWindowsLocked()
+			// Only a live user red-X still has this registered; programmatic closers
+			// nil s.browserLogin first and clean up themselves. Guarding here stops a
+			// stale close event from wiping a replacement popup's state.
+			userClosed := s.browserLogin == bl
+			if userClosed {
+				s.browserLogin = nil
+				s.restoreHiddenWindowsLocked()
+			}
 			s.mu.Unlock()
+			if userClosed {
+				s.app.Event.Emit(EventBrowserLoginCancel)
+			}
 		})
-		s.centerWhenReady(s.browserLogin)
+		s.centerOnCursorScreen(s.browserLogin)
 		return
 	}
 	if uri != "" {
 		s.browserLogin.SetURL("/#/dialog/browser-login?uri=" + url.QueryEscape(uri))
 	}
+	s.centerOnCursorScreen(s.browserLogin)
 	s.browserLogin.Show()
 	s.browserLogin.Focus()
-	s.centerWhenReady(s.browserLogin)
 }
 
 // BrowserLoginWindow returns the live SSO popup, or nil. While non-nil it is the
@@ -238,6 +239,15 @@ func (s *WindowManager) CloseBrowserLogin() {
 	s.mu.Lock()
 	w := s.browserLogin
 	s.browserLogin = nil
+	// The WindowClosing hook no-ops on a programmatic close, so restore here —
+	// but only if a popup was actually open. The frontend calls this even when no
+	// popup was ever shown (e.g. resetDialog() after an early RequestExtend failure,
+	// or connection.ts's catch path), and hiddenForLogin is shared with
+	// OpenInstallProgress, so an unconditional restore could re-show windows a
+	// still-running install-progress is hiding.
+	if w != nil {
+		s.restoreHiddenWindowsLocked()
+	}
 	s.mu.Unlock()
 	if w != nil {
 		w.Close()
@@ -279,6 +289,35 @@ func (s *WindowManager) CloseSessionExpiration() {
 	}
 }
 
+// CloseRenewFlow tears down the SSO session-renewal UI in a single call: it
+// closes the browser-login popup and the session-expiration window together.
+func (s *WindowManager) CloseRenewFlow() {
+	s.mu.Lock()
+	bl := s.browserLogin
+	se := s.sessionExpiration
+	s.browserLogin = nil
+	s.sessionExpiration = nil
+	if se != nil {
+		kept := s.hiddenForLogin[:0]
+		for _, w := range s.hiddenForLogin {
+			if w != se {
+				kept = append(kept, w)
+			}
+		}
+		s.hiddenForLogin = kept
+	}
+	s.restoreHiddenWindowsLocked()
+	s.mu.Unlock()
+
+	// Close after unlock so the re-entrant handlers can take s.mu.
+	if bl != nil {
+		bl.Close()
+	}
+	if se != nil {
+		se.Close()
+	}
+}
+
 // OpenInstallProgress shows the install-progress window and hides the rest for the duration
 // (restored on close). It owns its own result polling since the daemon restarts mid-install.
 func (s *WindowManager) OpenInstallProgress(version string) {

client/ui/tray.go

@@ -30,9 +30,10 @@ const (
 
 	statusError = "Error"
 
-	urlGitHubRepo     = "https://github.com/netbirdio/netbird"
-	urlGitHubReleases = "https://github.com/netbirdio/netbird/releases/latest"
-	urlDocs           = "https://docs.netbird.io"
+	quitDownTimeout = 5 * time.Second
+
+	urlGitHubRepo = "https://github.com/netbirdio/netbird"
+	urlDocs       = "https://docs.netbird.io"
 )
 
 // TrayServices bundles the services the tray menu needs, grouped so NewTray
@@ -315,8 +316,7 @@ func (t *Tray) relayoutMenu() {
 		if sessionDeadline.IsZero() {
 			t.sessionExpiresItem.SetHidden(true)
 		} else {
-			remaining := t.formatSessionRemaining(time.Until(sessionDeadline))
-			t.sessionExpiresItem.SetLabel(t.loc.T("tray.session.expiresIn", "remaining", remaining))
+			t.sessionExpiresItem.SetLabel(t.sessionRowLabel(sessionDeadline))
 			t.sessionExpiresItem.SetHidden(false)
 		}
 	}
@@ -446,11 +446,28 @@ func (t *Tray) buildMenu() *application.Menu {
 	menu.AddSeparator()
 	menu.Add(t.loc.T("tray.menu.quit")).
 		SetAccelerator("CmdOrCtrl+Q").
-		OnClick(func(*application.Context) { t.app.Quit() })
+		OnClick(func(*application.Context) { t.handleQuit() })
 
 	return menu
 }
 
+func (t *Tray) handleQuit() {
+	t.profileMu.Lock()
+	if t.switchCancel != nil {
+		t.switchCancel()
+		t.switchCancel = nil
+	}
+	t.profileMu.Unlock()
+	t.svc.DaemonFeed.CancelProfileSwitch()
+
+	ctx, cancel := context.WithTimeout(context.Background(), quitDownTimeout)
+	defer cancel()
+	if err := t.svc.Connection.Down(ctx); err != nil {
+		log.Errorf("disconnect on quit: %v", err)
+	}
+	t.app.Quit()
+}
+
 // handleConnect receives the clicked item from the buildMenu closure —
 // t.upItem is menuMu-guarded and must not be read here.
 func (t *Tray) handleConnect(upItem *application.MenuItem) {

client/ui/tray_session.go

@@ -64,11 +64,42 @@ func (t *Tray) applySessionExpiry(deadline *time.Time, connected bool) bool {
 	return changed
 }
 
-// runSessionExpiryTicker recomputes the "Expires in …" row label every 30s. Runs until process exit.
+// runSessionExpiryTicker recomputes the "Expires in …" row label until process exit.
+// The interval scales with the remaining time: coarse when the deadline is far off,
+// down to 10s in the final two minutes so the label doesn't lag the ceiling-rounded
+// countdown near expiry. The cached deadline is re-read every iteration, so an extend
+// or reconnect that moves it is picked up on the next tick.
 func (t *Tray) runSessionExpiryTicker() {
-	tk := time.NewTicker(30 * time.Second)
-	for range tk.C {
+	tm := time.NewTimer(sessionRefreshInterval(t.sessionRemaining()))
+	defer tm.Stop()
+	for range tm.C {
 		t.refreshSessionExpiresLabel()
+		tm.Reset(sessionRefreshInterval(t.sessionRemaining()))
+	}
+}
+
+// sessionRemaining returns the time left on the cached SSO deadline, or 0 when unknown.
+func (t *Tray) sessionRemaining() time.Duration {
+	t.sessionMu.Lock()
+	deadline := t.sessionExpiresAt
+	t.sessionMu.Unlock()
+	if deadline.IsZero() {
+		return 0
+	}
+	return time.Until(deadline)
+}
+
+// sessionRefreshInterval picks how long to wait before the next label recompute.
+func sessionRefreshInterval(remaining time.Duration) time.Duration {
+	switch {
+	case remaining <= 0:
+		return 30 * time.Second
+	case remaining <= 2*time.Minute:
+		return 10 * time.Second
+	case remaining <= time.Hour:
+		return 30 * time.Second
+	default:
+		return time.Minute
 	}
 }
 
@@ -87,30 +118,39 @@ func (t *Tray) refreshSessionExpiresLabel() {
 	if deadline.IsZero() {
 		return
 	}
-	remaining := t.formatSessionRemaining(time.Until(deadline))
-	item.SetLabel(t.loc.T("tray.session.expiresIn", "remaining", remaining))
+	item.SetLabel(t.sessionRowLabel(deadline))
+}
+
+func (t *Tray) sessionRowLabel(deadline time.Time) string {
+	remaining := time.Until(deadline)
+	if remaining <= 0 {
+		return t.loc.T("tray.status.sessionExpired")
+	}
+	return t.loc.T("tray.session.expiresIn", "remaining", t.formatSessionRemaining(remaining))
 }
 
 // formatSessionRemaining renders d as a localised long-form string picking the largest non-zero unit.
+// Each unit is rounded up so the label never claims less time than actually remains, matching the
+// upper-bound sense of the sub-minute "less than a minute" fragment.
 // Singular/plural keys are split per language for proper translation.
 func (t *Tray) formatSessionRemaining(d time.Duration) string {
 	switch {
 	case d < time.Minute:
 		return t.loc.T("tray.session.unit.lessThanMinute")
-	case d < time.Hour:
-		m := int(d / time.Minute)
+	case d <= 59*time.Minute:
+		m := ceilDiv(d, time.Minute)
 		if m == 1 {
 			return t.loc.T("tray.session.unit.minute")
 		}
 		return t.loc.T("tray.session.unit.minutes", "count", strconv.Itoa(m))
-	case d < 24*time.Hour:
-		h := int((d + 30*time.Minute) / time.Hour)
+	case d <= 23*time.Hour:
+		h := ceilDiv(d, time.Hour)
 		if h == 1 {
 			return t.loc.T("tray.session.unit.hour")
 		}
 		return t.loc.T("tray.session.unit.hours", "count", strconv.Itoa(h))
 	default:
-		days := int((d + 12*time.Hour) / (24 * time.Hour))
+		days := ceilDiv(d, 24*time.Hour)
 		if days == 1 {
 			return t.loc.T("tray.session.unit.day")
 		}
@@ -118,6 +158,11 @@ func (t *Tray) formatSessionRemaining(d time.Duration) string {
 	}
 }
 
+// ceilDiv divides d by unit rounding up, assuming d > 0.
+func ceilDiv(d, unit time.Duration) int {
+	return int((d + unit - time.Nanosecond) / unit)
+}
+
 // registerSessionWarningCategory wires the OS notification category and response handler for the expiry warning.
 // Errors are swallowed since the worst case is a plain notification without buttons.
 func (t *Tray) registerSessionWarningCategory() {
@@ -252,11 +297,9 @@ func (t *Tray) openSessionExpiration() {
 }
 
 // openSessionExtendFlow opens the SessionExpiration window seeded with the cached deadline's remaining time,
-// for the "Expires in …" tray row. No-ops when the deadline is unknown or elapsed.
+// for the "Expires in …" tray row. Once the deadline has elapsed the row reads "Session expired" and the
+// click routes to the login flow instead. No-op when the deadline is unknown.
 func (t *Tray) openSessionExtendFlow() {
-	if t.svc.WindowManager == nil {
-		return
-	}
 	t.sessionMu.Lock()
 	deadline := t.sessionExpiresAt
 	t.sessionMu.Unlock()
@@ -265,6 +308,14 @@ 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()
+		}
+		return
+	}
+	if t.svc.WindowManager == nil {
 		return
 	}
 	t.svc.WindowManager.OpenSessionExpiration(seconds)

client/ui/tray_update.go

@@ -13,6 +13,7 @@ import (
 
 	"github.com/netbirdio/netbird/client/ui/services"
 	"github.com/netbirdio/netbird/client/ui/updater"
+	"github.com/netbirdio/netbird/version"
 )
 
 // trayUpdater owns the tray UI that reacts to auto-update. Composed inside Tray.
@@ -76,15 +77,15 @@ func (u *trayUpdater) applyLanguage() {
 	u.refreshMenuItem(state)
 }
 
-// handleClick opens the GitHub releases page when not Enforced, otherwise shows
-// the progress page and asks the daemon to start the installer.
+// handleClick opens the installer download link when not Enforced, otherwise
+// shows the progress page and asks the daemon to start the installer.
 func (u *trayUpdater) handleClick() {
 	u.mu.Lock()
 	state := u.state
 	u.mu.Unlock()
 
 	if !state.Enforced {
-		_ = u.app.Browser.OpenURL(urlGitHubReleases)
+		_ = u.app.Browser.OpenURL(version.DownloadUrl())
 		return
 	}
 

combined/cmd/admin.go

@@ -0,0 +1,151 @@
+package cmd
+
+import (
+	"context"
+	"fmt"
+
+	"github.com/dexidp/dex/storage"
+	log "github.com/sirupsen/logrus"
+	"github.com/spf13/cobra"
+
+	"github.com/netbirdio/netbird/formatter/hook"
+	admincmd "github.com/netbirdio/netbird/management/cmd/admin"
+	tokencmd "github.com/netbirdio/netbird/management/cmd/token"
+	nbconfig "github.com/netbirdio/netbird/management/internals/server/config"
+	"github.com/netbirdio/netbird/management/server/activity"
+	activitystore "github.com/netbirdio/netbird/management/server/activity/store"
+	"github.com/netbirdio/netbird/management/server/store"
+	"github.com/netbirdio/netbird/management/server/types"
+	"github.com/netbirdio/netbird/util"
+)
+
+// newAdminCommands creates the admin command tree with combined-specific resource openers.
+func newAdminCommands() *cobra.Command {
+	return admincmd.NewCommands(admincmd.Openers{
+		Resources: withAdminResources,
+		Store:     withAdminStoreOnly,
+		IDP:       withAdminIDPOnly,
+	})
+}
+
+func newLegacyTokenCommand() *cobra.Command {
+	cmd := tokencmd.NewCommands(tokencmd.StoreOpener(withAdminStoreOnly))
+	cmd.Deprecated = "use 'admin token' instead"
+	return cmd
+}
+
+// withAdminResources loads the combined YAML config, initializes stores, and calls fn.
+func withAdminResources(cmd *cobra.Command, fn func(ctx context.Context, resources admincmd.Resources) error) error {
+	return withAdminConfig(cmd, func(ctx context.Context, cfg *CombinedConfig) error {
+		mgmtConfig, err := adminManagementConfig(cfg)
+		if err != nil {
+			return err
+		}
+
+		managementStore, err := openAdminStore(ctx, cfg)
+		if err != nil {
+			return err
+		}
+		defer admincmd.CloseStore(ctx, managementStore)
+
+		idpStorage, idpStorageFile, err := admincmd.OpenIDPStorage(mgmtConfig)
+		if err != nil {
+			return err
+		}
+		defer admincmd.CloseIDPStorage(idpStorage)
+
+		eventStore, esErr := openAdminEventStore(ctx, cfg, mgmtConfig)
+		if esErr != nil {
+			_, _ = fmt.Fprintf(cmd.ErrOrStderr(), "Warning: audit events will not be recorded: %v\n", esErr)
+		}
+		if eventStore != nil {
+			defer func() {
+				if err := eventStore.Close(ctx); err != nil {
+					log.Debugf("close activity event store: %v", err)
+				}
+			}()
+		}
+
+		return fn(ctx, admincmd.Resources{Store: managementStore, IDPStorage: idpStorage, IDPStorageFile: idpStorageFile, EventStore: eventStore})
+	})
+}
+
+// withAdminStoreOnly opens only the management store for admin subcommands that do not
+// need embedded IdP storage.
+func withAdminStoreOnly(cmd *cobra.Command, fn func(ctx context.Context, s store.Store) error) error {
+	return withAdminConfig(cmd, func(ctx context.Context, cfg *CombinedConfig) error {
+		managementStore, err := openAdminStore(ctx, cfg)
+		if err != nil {
+			return err
+		}
+		defer admincmd.CloseStore(ctx, managementStore)
+
+		return fn(ctx, managementStore)
+	})
+}
+
+func withAdminIDPOnly(cmd *cobra.Command, fn func(ctx context.Context, idpStorage storage.Storage, storageFile string) error) error {
+	return withAdminConfig(cmd, func(ctx context.Context, cfg *CombinedConfig) error {
+		mgmtConfig, err := adminManagementConfig(cfg)
+		if err != nil {
+			return err
+		}
+		idpStorage, idpStorageFile, err := admincmd.OpenIDPStorage(mgmtConfig)
+		if err != nil {
+			return err
+		}
+		defer admincmd.CloseIDPStorage(idpStorage)
+
+		return fn(ctx, idpStorage, idpStorageFile)
+	})
+}
+
+func withAdminConfig(cmd *cobra.Command, fn func(ctx context.Context, cfg *CombinedConfig) error) error {
+	if err := util.InitLog("error", "console"); err != nil {
+		return fmt.Errorf("init log: %w", err)
+	}
+
+	ctx := context.WithValue(cmd.Context(), hook.ExecutionContextKey, hook.SystemSource) //nolint:staticcheck
+
+	cfg, err := LoadConfig(configPath)
+	if err != nil {
+		return fmt.Errorf("load config: %w", err)
+	}
+	cfg.ApplyAdminDefaults()
+	applyServerStoreEnv(cfg.Server.Store)
+
+	return fn(ctx, cfg)
+}
+
+func adminManagementConfig(cfg *CombinedConfig) (*nbconfig.Config, error) {
+	mgmtConfig, err := cfg.ToManagementConfig()
+	if err != nil {
+		return nil, fmt.Errorf("create management config: %w", err)
+	}
+	return mgmtConfig, nil
+}
+
+func openAdminStore(ctx context.Context, cfg *CombinedConfig) (store.Store, error) {
+	managementStore, err := store.NewStore(ctx, types.Engine(cfg.Management.Store.Engine), cfg.Management.DataDir, nil, true)
+	if err != nil {
+		return nil, fmt.Errorf("create store: %w", err)
+	}
+	return managementStore, nil
+}
+
+func openAdminEventStore(ctx context.Context, cfg *CombinedConfig, config *nbconfig.Config) (activity.Store, error) {
+	if config.DataStoreEncryptionKey == "" {
+		return nil, fmt.Errorf("data store encryption key is not configured")
+	}
+	if err := applyActivityStoreEnv(cfg.Server.ActivityStore); err != nil {
+		return nil, fmt.Errorf("configure activity event store: %w", err)
+	}
+	eventStore, err := activitystore.NewSqlStore(ctx, config.Datadir, config.DataStoreEncryptionKey)
+	if err != nil {
+		return nil, fmt.Errorf("open activity event store: %w", err)
+	}
+	if eventStore == nil {
+		return nil, fmt.Errorf("open activity event store: returned nil store")
+	}
+	return eventStore, nil
+}

combined/cmd/admin_config_test.go

@@ -0,0 +1,47 @@
+package cmd
+
+import (
+	"context"
+	"os"
+	"testing"
+
+	"github.com/stretchr/testify/require"
+
+	nbconfig "github.com/netbirdio/netbird/management/internals/server/config"
+)
+
+func TestApplyAdminDefaultsCopiesServerStoreWithoutExposedAddress(t *testing.T) {
+	cfg := DefaultConfig()
+	cfg.Server.ExposedAddress = ""
+	cfg.Server.DataDir = "/srv/netbird"
+	cfg.Server.Store = StoreConfig{
+		Engine: "postgres",
+		DSN:    "postgres://user:pass@example.com/netbird",
+	}
+
+	cfg.ApplyAdminDefaults()
+
+	require.Equal(t, "/srv/netbird", cfg.Management.DataDir)
+	require.Equal(t, "postgres", cfg.Management.Store.Engine)
+	require.Equal(t, cfg.Server.Store.DSN, cfg.Management.Store.DSN)
+}
+
+func TestOpenAdminEventStoreMissingEncryptionKeyReturnsNilInterface(t *testing.T) {
+	eventStore, err := openAdminEventStore(context.Background(), &CombinedConfig{}, &nbconfig.Config{})
+	require.Error(t, err)
+	require.Contains(t, err.Error(), "encryption key")
+	require.Nil(t, eventStore)
+}
+
+func TestApplyServerStoreEnv(t *testing.T) {
+	t.Setenv("NB_STORE_ENGINE_POSTGRES_DSN", "")
+	t.Setenv("NB_STORE_ENGINE_MYSQL_DSN", "")
+	t.Setenv("NB_STORE_ENGINE_SQLITE_FILE", "")
+
+	applyServerStoreEnv(StoreConfig{Engine: "postgres", DSN: "postgres-dsn", File: "store.db"})
+	require.Equal(t, "postgres-dsn", os.Getenv("NB_STORE_ENGINE_POSTGRES_DSN"))
+	require.Equal(t, "store.db", os.Getenv("NB_STORE_ENGINE_SQLITE_FILE"))
+
+	applyServerStoreEnv(StoreConfig{Engine: "mysql", DSN: "mysql-dsn"})
+	require.Equal(t, "mysql-dsn", os.Getenv("NB_STORE_ENGINE_MYSQL_DSN"))
+}

combined/cmd/config.go

@@ -6,8 +6,7 @@ import (
 	"net"
 	"net/netip"
 	"os"
-	"path"
-	"path/filepath"
+	filePath "path/filepath"
 	"strings"
 	"time"
 
@@ -74,6 +73,9 @@ type ServerConfig struct {
 	ActivityStore           StoreConfig        `yaml:"activityStore"`
 	AuthStore               StoreConfig        `yaml:"authStore"`
 	ReverseProxy            ReverseProxyConfig `yaml:"reverseProxy"`
+
+	SupportedSyncMessageVersions           *int           `yaml:"supportedSyncMessageVersions,omitempty"`
+	PerAccountSupportedSyncMessageVersions map[string]int `yaml:"perAccountSupportedSyncMessageVersions,omitempty"`
 }
 
 // TLSConfig contains TLS/HTTPS settings
@@ -300,6 +302,19 @@ func (c *CombinedConfig) ApplySimplifiedDefaults() {
 	c.autoConfigureClientSettings(exposedProto, exposedHost, exposedHostPort, hasExternalStuns, hasExternalRelay, hasExternalSignal)
 }
 
+// ApplyAdminDefaults applies the management settings needed by admin commands even
+// when the full server config is invalid and ApplySimplifiedDefaults cannot run.
+func (c *CombinedConfig) ApplyAdminDefaults() {
+	if c.Management.DataDir == "" || c.Management.DataDir == "/var/lib/netbird/" {
+		c.Management.DataDir = c.Server.DataDir
+	}
+	if c.Management.Store.Engine == "" || c.Management.Store.Engine == "sqlite" {
+		if c.Server.Store.Engine != "" || c.Server.Store.File != "" || c.Server.Store.DSN != "" {
+			c.Management.Store = c.Server.Store
+		}
+	}
+}
+
 // applyRelayDefaults configures the relay service if no external relay is configured.
 func (c *CombinedConfig) applyRelayDefaults(exposedProto, exposedHostPort string, hasExternalRelay, hasExternalStuns bool) {
 	if hasExternalRelay {
@@ -577,11 +592,11 @@ func (c *CombinedConfig) buildEmbeddedIdPConfig(mgmt ManagementConfig) (*idp.Emb
 			return nil, fmt.Errorf("authStore.dsn is required when authStore.engine is postgres")
 		}
 	} else {
-		authStorageFile = path.Join(mgmt.DataDir, "idp.db")
+		authStorageFile = filePath.Join(mgmt.DataDir, "idp.db")
 		if c.Server.AuthStore.File != "" {
 			authStorageFile = c.Server.AuthStore.File
-			if !filepath.IsAbs(authStorageFile) {
-				authStorageFile = filepath.Join(mgmt.DataDir, authStorageFile)
+			if !filePath.IsAbs(authStorageFile) {
+				authStorageFile = filePath.Join(mgmt.DataDir, authStorageFile)
 			}
 		}
 	}
@@ -696,16 +711,18 @@ func (c *CombinedConfig) ToManagementConfig() (*nbconfig.Config, error) {
 	httpConfig.AuthCallbackURL = callbackURL + types.ProxyCallbackEndpointFull
 
 	return &nbconfig.Config{
-		Stuns:                  stuns,
-		Relay:                  relayConfig,
-		Signal:                 signalConfig,
-		Datadir:                mgmt.DataDir,
-		DataStoreEncryptionKey: mgmt.Store.EncryptionKey,
-		HttpConfig:             httpConfig,
-		StoreConfig:            storeConfig,
-		ReverseProxy:           reverseProxy,
-		DisableDefaultPolicy:   mgmt.DisableDefaultPolicy,
-		EmbeddedIdP:            embeddedIdP,
+		Stuns:                              stuns,
+		Relay:                              relayConfig,
+		Signal:                             signalConfig,
+		Datadir:                            mgmt.DataDir,
+		DataStoreEncryptionKey:             mgmt.Store.EncryptionKey,
+		HttpConfig:                         httpConfig,
+		StoreConfig:                        storeConfig,
+		ReverseProxy:                       reverseProxy,
+		DisableDefaultPolicy:               mgmt.DisableDefaultPolicy,
+		EmbeddedIdP:                        embeddedIdP,
+		HighestSupportedSyncMessageVersion: c.Server.SupportedSyncMessageVersions,
+		PerAccountHighestSupportedSyncMessageVersion: c.Server.PerAccountSupportedSyncMessageVersions,
 	}, nil
 }
 
@@ -729,7 +746,7 @@ func ApplyEmbeddedIdPConfig(ctx context.Context, cfg *nbconfig.Config, mgmtPort
 		cfg.EmbeddedIdP.Storage.Type = "sqlite3"
 	}
 	if cfg.EmbeddedIdP.Storage.Config.File == "" && cfg.Datadir != "" {
-		cfg.EmbeddedIdP.Storage.Config.File = path.Join(cfg.Datadir, "idp.db")
+		cfg.EmbeddedIdP.Storage.Config.File = filePath.Join(cfg.Datadir, "idp.db")
 	}
 
 	issuer := cfg.EmbeddedIdP.Issuer

combined/cmd/root.go

@@ -31,6 +31,7 @@ import (
 	relayServer "github.com/netbirdio/netbird/relay/server"
 	"github.com/netbirdio/netbird/relay/server/listener"
 	"github.com/netbirdio/netbird/relay/server/listener/ws"
+	syncgrpc "github.com/netbirdio/netbird/shared/management/grpc"
 	sharedMetrics "github.com/netbirdio/netbird/shared/metrics"
 	"github.com/netbirdio/netbird/shared/relay/auth"
 	"github.com/netbirdio/netbird/shared/signal/proto"
@@ -64,7 +65,8 @@ func init() {
 	rootCmd.PersistentFlags().StringVarP(&configPath, "config", "c", "", "path to YAML configuration file (required)")
 	_ = rootCmd.MarkPersistentFlagRequired("config")
 
-	rootCmd.AddCommand(newTokenCommands())
+	rootCmd.AddCommand(newAdminCommands())
+	rootCmd.AddCommand(newLegacyTokenCommand())
 }
 
 func RootCmd() *cobra.Command {
@@ -122,6 +124,37 @@ func execute(cmd *cobra.Command, _ []string) error {
 }
 
 // initializeConfig loads and validates the configuration, then initializes logging.
+func applyServerStoreEnv(storeConfig StoreConfig) {
+	if dsn := storeConfig.DSN; dsn != "" {
+		switch strings.ToLower(storeConfig.Engine) {
+		case "postgres":
+			os.Setenv("NB_STORE_ENGINE_POSTGRES_DSN", dsn)
+		case "mysql":
+			os.Setenv("NB_STORE_ENGINE_MYSQL_DSN", dsn)
+		}
+	}
+	if file := storeConfig.File; file != "" {
+		os.Setenv("NB_STORE_ENGINE_SQLITE_FILE", file)
+	}
+}
+
+func applyActivityStoreEnv(storeConfig StoreConfig) error {
+	if engine := storeConfig.Engine; engine != "" {
+		engineLower := strings.ToLower(engine)
+		if engineLower == "postgres" && storeConfig.DSN == "" {
+			return fmt.Errorf("activityStore.dsn is required when activityStore.engine is postgres")
+		}
+		os.Setenv("NB_ACTIVITY_EVENT_STORE_ENGINE", engineLower)
+		if dsn := storeConfig.DSN; dsn != "" {
+			os.Setenv("NB_ACTIVITY_EVENT_POSTGRES_DSN", dsn)
+		}
+	}
+	if file := storeConfig.File; file != "" {
+		os.Setenv("NB_ACTIVITY_EVENT_SQLITE_FILE", file)
+	}
+	return nil
+}
+
 func initializeConfig() error {
 	var err error
 	config, err = LoadConfig(configPath)
@@ -137,30 +170,10 @@ func initializeConfig() error {
 		return fmt.Errorf("failed to initialize log: %w", err)
 	}
 
-	if dsn := config.Server.Store.DSN; dsn != "" {
-		switch strings.ToLower(config.Server.Store.Engine) {
-		case "postgres":
-			os.Setenv("NB_STORE_ENGINE_POSTGRES_DSN", dsn)
-		case "mysql":
-			os.Setenv("NB_STORE_ENGINE_MYSQL_DSN", dsn)
-		}
-	}
-	if file := config.Server.Store.File; file != "" {
-		os.Setenv("NB_STORE_ENGINE_SQLITE_FILE", file)
-	}
+	applyServerStoreEnv(config.Server.Store)
 
-	if engine := config.Server.ActivityStore.Engine; engine != "" {
-		engineLower := strings.ToLower(engine)
-		if engineLower == "postgres" && config.Server.ActivityStore.DSN == "" {
-			return fmt.Errorf("activityStore.dsn is required when activityStore.engine is postgres")
-		}
-		os.Setenv("NB_ACTIVITY_EVENT_STORE_ENGINE", engineLower)
-		if dsn := config.Server.ActivityStore.DSN; dsn != "" {
-			os.Setenv("NB_ACTIVITY_EVENT_POSTGRES_DSN", dsn)
-		}
-	}
-	if file := config.Server.ActivityStore.File; file != "" {
-		os.Setenv("NB_ACTIVITY_EVENT_SQLITE_FILE", file)
+	if err := applyActivityStoreEnv(config.Server.ActivityStore); err != nil {
+		return err
 	}
 
 	log.Infof("Starting combined NetBird server")
@@ -505,6 +518,16 @@ func createManagementServer(cfg *CombinedConfig, mgmtConfig *nbconfig.Config) (m
 	}
 	mgmtPort, _ := strconv.Atoi(portStr)
 
+	if err := syncgrpc.ValidateSyncMessageVersion(mgmtConfig.HighestSupportedSyncMessageVersion); err != nil {
+		return nil, err
+	}
+
+	for accountId, version := range mgmtConfig.PerAccountHighestSupportedSyncMessageVersion {
+		if err := syncgrpc.ValidateSyncMessageVersion(&version); err != nil {
+			return nil, fmt.Errorf("unrecognized sync message version in perAccountSupportedSyncMessageVersions for account %s %w", accountId, err)
+		}
+	}
+
 	mgmtSrv := newServer(
 		&mgmtServer.Config{
 			NbConfig:                mgmtConfig,

combined/cmd/token.go

@@ -1,63 +0,0 @@
-package cmd
-
-import (
-	"context"
-	"fmt"
-	"os"
-	"strings"
-
-	log "github.com/sirupsen/logrus"
-	"github.com/spf13/cobra"
-
-	"github.com/netbirdio/netbird/formatter/hook"
-	tokencmd "github.com/netbirdio/netbird/management/cmd/token"
-	"github.com/netbirdio/netbird/management/server/store"
-	"github.com/netbirdio/netbird/management/server/types"
-	"github.com/netbirdio/netbird/util"
-)
-
-// newTokenCommands creates the token command tree with combined-specific store opener.
-func newTokenCommands() *cobra.Command {
-	return tokencmd.NewCommands(withTokenStore)
-}
-
-// withTokenStore loads the combined YAML config, initializes the store, and calls fn.
-func withTokenStore(cmd *cobra.Command, fn func(ctx context.Context, s store.Store) error) error {
-	if err := util.InitLog("error", "console"); err != nil {
-		return fmt.Errorf("init log: %w", err)
-	}
-
-	ctx := context.WithValue(cmd.Context(), hook.ExecutionContextKey, hook.SystemSource) //nolint:staticcheck
-
-	cfg, err := LoadConfig(configPath)
-	if err != nil {
-		return fmt.Errorf("load config: %w", err)
-	}
-
-	if dsn := cfg.Server.Store.DSN; dsn != "" {
-		switch strings.ToLower(cfg.Server.Store.Engine) {
-		case "postgres":
-			os.Setenv("NB_STORE_ENGINE_POSTGRES_DSN", dsn)
-		case "mysql":
-			os.Setenv("NB_STORE_ENGINE_MYSQL_DSN", dsn)
-		}
-	}
-	if file := cfg.Server.Store.File; file != "" {
-		os.Setenv("NB_STORE_ENGINE_SQLITE_FILE", file)
-	}
-
-	datadir := cfg.Management.DataDir
-	engine := types.Engine(cfg.Management.Store.Engine)
-
-	s, err := store.NewStore(ctx, engine, datadir, nil, true)
-	if err != nil {
-		return fmt.Errorf("create store: %w", err)
-	}
-	defer func() {
-		if err := s.Close(ctx); err != nil {
-			log.Debugf("close store: %v", err)
-		}
-	}()
-
-	return fn(ctx, s)
-}

dns/nameserver.go

@@ -53,6 +53,7 @@ type NameServerGroup struct {
 	ID string `gorm:"primaryKey"`
 	// AccountID is a reference to Account that this object belongs
 	AccountID string `gorm:"index"`
+	PublicID  string `json:"-"`
 	// Name group name
 	Name string
 	// Description group description

docs/agent-networks/01-end-to-end-flows.md

@@ -109,7 +109,7 @@ sequenceDiagram
         Chk->>Inj: continue
         Inj->>Inj: inject NetBird identity headers per provider config
         Inj->>Grd: continue
-        Grd->>Grd: enforce model allowlist
+        Grd->>Grd: enforce per-provider allowlist (fail-closed backstop)
         Grd->>Up: forward (over WireGuard)
         Up-->>Resp: response (JSON or SSE stream)
         Resp->>Resp: parse usage tokens, completion
@@ -135,6 +135,21 @@ sequenceDiagram
   (`redact_pii = settings.RedactPii`). Phones, emails, credit cards,
   PII names — see `redact.go` for the full set. See
   [`modules/31-proxy-middleware-builtin.md`](modules/31-proxy-middleware-builtin.md).
+- The model allowlist is enforced in TWO places. `CheckLLMPolicyLimits`
+  is authoritative: it resolves the policy that governs this
+  (provider, caller-groups) and denies (`deny_code = llm_policy.model_blocked`)
+  when no applicable policy permits the model — so an allowlist scoped to
+  one group/provider never leaks to another, and an un-guardrailed policy
+  is genuinely unrestricted. `llm_guardrail` is a per-provider fail-closed
+  backstop: it only carries an allowlist for a provider every authorising
+  policy restricts, and blocks unknown/undetermined models even when
+  management is unreachable. Because that backstop allowlist is the UNION
+  of every restricting policy's models, per-group narrowing lives only in
+  the authoritative check: during a `CheckLLMPolicyLimits` outage
+  `llm_limit_check` fails open, so a caller can reach any model in the
+  provider's union — a group scoped to model A could reach model B if
+  another group restricts the same provider to B. This is the documented
+  fail-open trade-off; a future flag may switch it to fail-closed.
 - SSE streaming requires special handling on the response side; the
   parser must handle partial chunks without buffering the whole
   stream. See [`modules/32-proxy-llm-parsers.md`](modules/32-proxy-llm-parsers.md).

docs/agent-networks/modules/21-management-agentnetwork.md

@@ -122,7 +122,7 @@ At request time the path is independent: the proxy calls `SelectPolicyForRequest
   | on_request | 1 | `llm_router` | `{"providers":[{id, models[], upstream_*, auth_header_*, allowed_group_ids[]}]}` | **true** |
   | on_request | 2 | `llm_limit_check` | `{}` | – |
   | on_request | 3 | `llm_identity_inject` | `{"providers":[{provider_id, header_pair?, json_metadata?, extra_headers?}]}` | **true** |
-  | on_request | 4 | `llm_guardrail` | `{"model_allowlist"?, "prompt_capture":{enabled,redact_pii}}` | – |
+  | on_request | 4 | `llm_guardrail` | `{"provider_allowlists"?: {providerID: []model}, "prompt_capture":{enabled,redact_pii}}` | – |
   | on_response | 5 | `llm_limit_record` | `{}` (runs LAST at runtime) | – |
   | on_response | 6 | `cost_meter` | `{}` | – |
   | on_response | 7 | `llm_response_parser` | `{"capture_completion": <bool>, "redact_pii"?: true}` | – |

docs/agent-networks/modules/31-proxy-middleware-builtin.md

@@ -244,7 +244,7 @@ no mocks. Tests: `TestChain_AllowPath_StampsAttributionAndRecordsCounter`
 | `llm_router` | `{providers: [{id, models, upstream_scheme, upstream_host, upstream_path?, auth_header_name, auth_header_value, allowed_group_ids}]}` |
 | `llm_limit_check` | `{}` — pulls `MgmtClient` from `FactoryContext` |
 | `llm_identity_inject` | `{providers: [{provider_id, header_pair?|json_metadata?, extra_headers?}]}` |
-| `llm_guardrail` | `{model_allowlist: []string, prompt_capture: {enabled, redact_pii}}` |
+| `llm_guardrail` | `{provider_allowlists: {providerID: []string}, prompt_capture: {enabled, redact_pii}}` — allowlist keyed by resolved provider id; a provider absent from the map is unrestricted (fail-closed backstop; authoritative per-policy/group check is management's `CheckLLMPolicyLimits`) |
 | `llm_response_parser` | `{redact_pii?, capture_completion?: *bool}` |
 | `cost_meter` | `{pricing_path?}` (basename inside data-dir; defaults `pricing.yaml`) |
 | `llm_limit_record` | `{}` — same pattern as `llm_limit_check` |

docs/io.netbird.client.plist

@@ -66,6 +66,9 @@
     <key>disableAutoConnect</key>
     <false/>
 
+    <key>disableAutostart</key>
+    <false/>
+
     <key>disableClientRoutes</key>
     <false/>
 

docs/netbird-macos.mobileconfig

@@ -103,6 +103,8 @@
                                 <!--
                                 <key>disableAutoConnect</key>
                                 <false/>
+                                <key>disableAutostart</key>
+                                <false/>
                                 <key>disableClientRoutes</key>
                                 <false/>
                                 <key>disableServerRoutes</key>

docs/netbird-macos.sh

@@ -58,6 +58,7 @@ preSharedKey="$NULL"                       # secret; redacted in log
 allowServerSSH='true'
 blockInbound="$NULL"
 disableAutoConnect="$NULL"
+disableAutostart="$NULL"
 disableClientRoutes="$NULL"
 disableServerRoutes="$NULL"
 disableMetricsCollection="$NULL"
@@ -155,6 +156,7 @@ main() {
   is_set "$allowServerSSH"            && emit_bool    allowServerSSH            "$allowServerSSH"
   is_set "$blockInbound"              && emit_bool    blockInbound              "$blockInbound"
   is_set "$disableAutoConnect"        && emit_bool    disableAutoConnect        "$disableAutoConnect"
+  is_set "$disableAutostart"          && emit_bool    disableAutostart          "$disableAutostart"
   is_set "$disableClientRoutes"       && emit_bool    disableClientRoutes       "$disableClientRoutes"
   is_set "$disableServerRoutes"       && emit_bool    disableServerRoutes       "$disableServerRoutes"
   is_set "$disableMetricsCollection"  && emit_bool    disableMetricsCollection  "$disableMetricsCollection"

docs/netbird.adml

@@ -24,6 +24,9 @@
       <string id="DisableAutoConnect_Name">Disable auto-connect</string>
       <string id="DisableAutoConnect_Help">When enabled, the NetBird tunnel does not auto-connect at daemon startup. Equivalent to --disable-auto-connect.</string>
 
+      <string id="DisableAutostart_Name">Disable autostart</string>
+      <string id="DisableAutostart_Help">When enabled, the NetBird GUI is prevented from registering itself as an OS autostart entry on fresh installs, and any existing OS autostart entry registration is removed on the next GUI launch (Windows Registry Run key, macOS Login Item, Linux .desktop). Once the admin lifts the policy, the setting stays off until the user re-enables it in Settings.</string>
+
       <string id="DisableClientRoutes_Name">Disable client routes</string>
       <string id="DisableClientRoutes_Help">When enabled, this client will not consume routes advertised by routing peers. Equivalent to --disable-client-routes.</string>
 

docs/netbird.admx

@@ -64,6 +64,18 @@
       <disabledValue><decimal value="0" /></disabledValue>
     </policy>
 
+    <policy name="DisableAutostart"
+            class="Machine"
+            displayName="$(string.DisableAutostart_Name)"
+            explainText="$(string.DisableAutostart_Help)"
+            key="Software\Policies\NetBird"
+            valueName="DisableAutostart">
+      <parentCategory ref="NetBird" />
+      <supportedOn ref="SUPPORTED_NetBird_All" />
+      <enabledValue><decimal value="1" /></enabledValue>
+      <disabledValue><decimal value="0" /></disabledValue>
+    </policy>
+
     <policy name="DisableClientRoutes"
             class="Machine"
             displayName="$(string.DisableClientRoutes_Name)"

e2e/agentnetwork/chat_test.go

@@ -9,25 +9,234 @@ import (
 	"testing"
 	"time"
 
+	"github.com/stretchr/testify/assert"
 	"github.com/stretchr/testify/require"
+	"gorm.io/driver/sqlite"
+	"gorm.io/gorm"
 
 	"github.com/netbirdio/netbird/e2e/harness"
 	"github.com/netbirdio/netbird/shared/management/http/api"
 )
 
+// per1k is a model's published USD rates per 1k tokens. read is the prompt-cache read rate
+// (OpenAI: the cached-input discount rate); write is the cache-creation rate where one exists.
+type per1k struct{ in, out, read, write float64 }
+
+// publishedPer1k hardcodes the vendors' PUBLISHED rates for the models the live matrix can drive,
+// keyed by the normalized model id the proxy stamps. Deliberately independent of the proxy's
+// pricing table so a wrong embedded rate or a broken normalization fails the run.
+var publishedPer1k = map[string]per1k{
+	"gpt-4o-mini":                 {0.00015, 0.0006, 0.000075, 0},
+	"gpt-4o":                      {0.0025, 0.01, 0.00125, 0},
+	"claude-haiku-4-5":            {0.001, 0.005, 0.0001, 0.00125},
+	"claude-sonnet-4-5":           {0.003, 0.015, 0.0003, 0.00375},
+	"claude-sonnet-4-6":           {0.003, 0.015, 0.0003, 0.00375},
+	"kimi-k3":                     {0.003, 0.015, 0.0003, 0.003}, // no published write rate: bills at the input rate
+	"anthropic.claude-haiku-4-5":  {0.001, 0.005, 0.0001, 0.00125},
+	"anthropic.claude-sonnet-4-5": {0.003, 0.015, 0.0003, 0.00375},
+	"anthropic.claude-sonnet-4-6": {0.003, 0.015, 0.0003, 0.00375},
+}
+
+// rawCostVerificationSQL is the operator-facing double-check, run straight against the management
+// sqlite store: recompute each usage row's expected total and cache cost from its own persisted
+// token buckets and hardcoded published rates. OpenAI counts cached tokens as a subset of input;
+// Anthropic-shape providers count cache buckets additively.
+const rawCostVerificationSQL = `
+WITH rates(model, in_rate, out_rate, read_rate, write_rate) AS (
+  VALUES
+    ('gpt-4o-mini',                 0.00015, 0.0006, 0.000075, 0.0),
+    ('gpt-4o',                      0.0025,  0.01,   0.00125,  0.0),
+    ('claude-haiku-4-5',            0.001,   0.005,  0.0001,   0.00125),
+    ('claude-sonnet-4-5',           0.003,   0.015,  0.0003,   0.00375),
+    ('claude-sonnet-4-6',           0.003,   0.015,  0.0003,   0.00375),
+    ('kimi-k3',                     0.003,   0.015,  0.0003,   0.003),
+    ('anthropic.claude-haiku-4-5',  0.001,   0.005,  0.0001,   0.00125),
+    ('anthropic.claude-sonnet-4-5', 0.003,   0.015,  0.0003,   0.00375),
+    ('anthropic.claude-sonnet-4-6', 0.003,   0.015,  0.0003,   0.00375)
+)
+SELECT
+  u.provider,
+  u.model,
+  u.input_tokens,
+  u.output_tokens,
+  u.cached_input_tokens,
+  u.cache_creation_tokens,
+  u.input_cost_usd,
+  u.cached_input_cost_usd,
+  u.cache_creation_cost_usd,
+  u.output_cost_usd,
+  -- No cost_usd / cache_cost_usd columns are stored: both are derived from the
+  -- four per-bucket columns above, exactly as the API renders them.
+  (u.input_cost_usd + u.cached_input_cost_usd + u.cache_creation_cost_usd + u.output_cost_usd) AS cost_usd,
+  (u.cached_input_cost_usd + u.cache_creation_cost_usd) AS cache_cost_usd,
+  CASE WHEN u.provider = 'openai' THEN
+    (u.input_tokens - MIN(u.cached_input_tokens, u.input_tokens))*r.in_rate/1000.0
+  ELSE
+    u.input_tokens*r.in_rate/1000.0
+  END AS expected_input,
+  CASE WHEN u.provider = 'openai' THEN
+    MIN(u.cached_input_tokens, u.input_tokens)*r.read_rate/1000.0
+  ELSE
+    u.cached_input_tokens*r.read_rate/1000.0
+  END AS expected_cached_input,
+  CASE WHEN u.provider = 'openai' THEN
+    0.0
+  ELSE
+    u.cache_creation_tokens*r.write_rate/1000.0
+  END AS expected_cache_creation,
+  u.output_tokens*r.out_rate/1000.0 AS expected_output,
+  CASE WHEN u.provider = 'openai' THEN
+    (u.input_tokens - MIN(u.cached_input_tokens, u.input_tokens))*r.in_rate/1000.0
+      + MIN(u.cached_input_tokens, u.input_tokens)*r.read_rate/1000.0
+      + u.output_tokens*r.out_rate/1000.0
+  ELSE
+    u.input_tokens*r.in_rate/1000.0 + u.cached_input_tokens*r.read_rate/1000.0
+      + u.cache_creation_tokens*r.write_rate/1000.0 + u.output_tokens*r.out_rate/1000.0
+  END AS expected_total,
+  CASE WHEN u.provider = 'openai' THEN
+    MIN(u.cached_input_tokens, u.input_tokens)*r.read_rate/1000.0
+  ELSE
+    u.cached_input_tokens*r.read_rate/1000.0 + u.cache_creation_tokens*r.write_rate/1000.0
+  END AS expected_cache
+FROM agent_network_request_usage u
+JOIN rates r ON r.model = u.model
+ORDER BY u.timestamp`
+
+// verifyUsageRowsSQL re-checks every persisted usage row directly in the management sqlite store,
+// bypassing the API path — the same audit an operator can run on a production store.db.
+func verifyUsageRowsSQL(t *testing.T, srv *harness.Combined) {
+	t.Helper()
+
+	dbPath, err := srv.SnapshotStoreDB(t.TempDir())
+	require.NoError(t, err, "snapshot management sqlite store")
+
+	db, err := gorm.Open(sqlite.Open(dbPath), &gorm.Config{})
+	require.NoError(t, err, "open store snapshot")
+	sqlDB, err := db.DB()
+	require.NoError(t, err)
+	defer func() { _ = sqlDB.Close() }()
+
+	rows, err := db.Raw(rawCostVerificationSQL).Rows()
+	require.NoError(t, err, "run raw cost verification query")
+	defer func() { _ = rows.Close() }()
+
+	verified := 0
+	for rows.Next() {
+		var provider, model string
+		var inTok, outTok, readTok, writeTok int64
+		var inCost, cachedInCost, cacheCreateCost, outCost, cost, cacheCost float64
+		var wantInput, wantCachedInput, wantCacheCreation, wantOutput, wantTotal, wantCache float64
+		require.NoError(t, rows.Scan(&provider, &model, &inTok, &outTok, &readTok, &writeTok,
+			&inCost, &cachedInCost, &cacheCreateCost, &outCost, &cost, &cacheCost,
+			&wantInput, &wantCachedInput, &wantCacheCreation, &wantOutput, &wantTotal, &wantCache), "scan usage row")
+		t.Logf("[sql] %s/%s: in=%d out=%d cache_read=%d cache_write=%d stored in/cached/create/out=$%.6f/$%.6f/$%.6f/$%.6f total=$%.6f cache=$%.6f expected total=$%.6f cache=$%.6f",
+			provider, model, inTok, outTok, readTok, writeTok,
+			inCost, cachedInCost, cacheCreateCost, outCost, cost, cacheCost, wantTotal, wantCache)
+		assert.InDeltaf(t, wantInput, inCost, 1e-6, "stored input_cost_usd for %s/%s must match the published-rate recompute", provider, model)
+		assert.InDeltaf(t, wantCachedInput, cachedInCost, 1e-6, "stored cached_input_cost_usd for %s/%s must match the published-rate recompute", provider, model)
+		assert.InDeltaf(t, wantCacheCreation, cacheCreateCost, 1e-6, "stored cache_creation_cost_usd for %s/%s must match the published-rate recompute", provider, model)
+		assert.InDeltaf(t, wantOutput, outCost, 1e-6, "stored output_cost_usd for %s/%s must match the published-rate recompute", provider, model)
+		assert.InDeltaf(t, wantTotal, cost, 1e-6, "derived cost_usd for %s/%s must match the published-rate recompute", provider, model)
+		assert.InDeltaf(t, wantCache, cacheCost, 1e-6, "derived cache_cost_usd for %s/%s must match the published-rate recompute", provider, model)
+		assert.InDeltaf(t, inCost+cachedInCost+cacheCreateCost+outCost, cost, 1e-9,
+			"stored buckets must sum to the derived cost_usd for %s/%s", provider, model)
+		verified++
+	}
+	require.NoError(t, rows.Err(), "iterate usage rows")
+	require.Positive(t, verified, "raw SQL check must cover at least one usage row")
+	t.Logf("[sql] verified %d usage rows in store.db against published rates", verified)
+
+	gwRows, err := db.Raw(`SELECT model,
+	  (input_cost_usd + cached_input_cost_usd + cache_creation_cost_usd + output_cost_usd) AS cost_usd
+	  FROM agent_network_request_usage WHERE model LIKE '%/%'`).Rows()
+	require.NoError(t, err, "query gateway-prefixed usage rows")
+	defer func() { _ = gwRows.Close() }()
+	for gwRows.Next() {
+		var model string
+		var cost float64
+		require.NoError(t, gwRows.Scan(&model, &cost), "scan gateway usage row")
+		t.Logf("[sql] gateway %s: stored=$%.6f (must be 0 — deliberately unpriced)", model, cost)
+		assert.Zerof(t, cost, "gateway-prefixed model %q must store cost 0, never a guessed rate", model)
+	}
+	require.NoError(t, gwRows.Err(), "iterate gateway usage rows")
+}
+
+// validateAccessLogCost recomputes a live access-log row's expected total and cache cost from the
+// published per-1k rates and the row's persisted token buckets, and asserts both stored values.
+// Gateway-prefixed model ids the proxy deliberately does not price must store cost 0.
+func validateAccessLogCost(t *testing.T, pc providerCase, row api.AgentNetworkAccessLog) {
+	t.Helper()
+	model := catalogModel(pc)
+	provider := ""
+	if row.Provider != nil {
+		provider = *row.Provider
+	}
+	t.Logf("[cost] %s: provider=%s model=%s in=%d out=%d total=%d cache_read=%d cache_write=%d cost=$%.6f cache_cost=$%.6f",
+		pc.name, provider, model, row.InputTokens, row.OutputTokens, row.TotalTokens,
+		row.CachedInputTokens, row.CacheCreationTokens, row.CostUsd, row.CacheCostUsd)
+
+	rates, known := publishedPer1k[model]
+	if !known {
+		if strings.Contains(model, "/") {
+			assert.Zerof(t, row.CostUsd, "gateway-prefixed model %q is not priced so the cost meter must skip (cost 0)", model)
+			return
+		}
+		t.Logf("[cost] %s: no published rate on file for model %q (env-overridden?); skipping cost validation", pc.name, model)
+		return
+	}
+
+	// input_tokens may legitimately be 0: Moonshot/Kimi reports fully cached prompts under the cache
+	// buckets only. Output and total must always be present on a priced row.
+	require.Positive(t, row.OutputTokens, "priced row must carry output tokens")
+	require.Positive(t, row.TotalTokens, "priced row must carry total tokens")
+
+	var wantInput, wantCachedInput, wantCacheCreation float64
+	if provider == "openai" {
+		cached := min(row.CachedInputTokens, row.InputTokens) // cached is a subset of input
+		wantInput = float64(row.InputTokens-cached) / 1000 * rates.in
+		wantCachedInput = float64(cached) / 1000 * rates.read
+		// OpenAI has no cache-write bucket; wantCacheCreation stays 0.
+	} else {
+		// Anthropic / Bedrock shape: cache buckets are additive to input_tokens.
+		wantInput = float64(row.InputTokens) / 1000 * rates.in
+		wantCachedInput = float64(row.CachedInputTokens) / 1000 * rates.read
+		wantCacheCreation = float64(row.CacheCreationTokens) / 1000 * rates.write
+	}
+	wantOutput := float64(row.OutputTokens) / 1000 * rates.out
+	wantCache := wantCachedInput + wantCacheCreation
+	wantTotal := wantInput + wantCache + wantOutput
+
+	t.Logf("[cost] %s: expecting input=$%.6f cached_input=$%.6f cache_creation=$%.6f output=$%.6f total=$%.6f cache=$%.6f from published rates",
+		pc.name, wantInput, wantCachedInput, wantCacheCreation, wantOutput, wantTotal, wantCache)
+	assert.InDeltaf(t, wantInput, row.InputCostUsd, 1e-6, "stored input_cost_usd for %s (%s)", pc.name, model)
+	assert.InDeltaf(t, wantCachedInput, row.CachedInputCostUsd, 1e-6, "stored cached_input_cost_usd for %s (%s)", pc.name, model)
+	assert.InDeltaf(t, wantCacheCreation, row.CacheCreationCostUsd, 1e-6, "stored cache_creation_cost_usd for %s (%s)", pc.name, model)
+	assert.InDeltaf(t, wantOutput, row.OutputCostUsd, 1e-6, "stored output_cost_usd for %s (%s)", pc.name, model)
+	assert.InDeltaf(t, wantTotal, row.CostUsd, 1e-6, "derived cost_usd for %s (%s)", pc.name, model)
+	assert.InDeltaf(t, wantCache, row.CacheCostUsd, 1e-6, "derived cache_cost_usd for %s (%s)", pc.name, model)
+
+	// The aggregates must be exactly the sum of the stored components, not an
+	// independently-computed figure that could drift from the breakdown.
+	assert.InDeltaf(t, row.InputCostUsd+row.CachedInputCostUsd+row.CacheCreationCostUsd+row.OutputCostUsd,
+		row.CostUsd, 1e-9, "stored buckets must sum to the derived cost_usd for %s (%s)", pc.name, model)
+	assert.InDeltaf(t, row.CachedInputCostUsd+row.CacheCreationCostUsd,
+		row.CacheCostUsd, 1e-9, "stored cache buckets must sum to the derived cache_cost_usd for %s (%s)", pc.name, model)
+}
+
 // providerCase is one entry in the live provider matrix. The same scenario runs
 // for every available provider; availability is keyed off env vars so the suite
 // covers whatever credentials are present (source ~/.llm-keys locally / set the
 // Actions secrets in CI).
 type providerCase struct {
-	name      string
-	catalogID string
-	upstream  string
-	apiKey    string
-	model     string // body model (chat/messages) or path model@version (vertex)
-	kind      string // harness.WireChat, harness.WireMessages, or harness.WireVertex
-	project   string // vertex only: GCP project for the rawPredict path
-	region    string // vertex only: GCP region for the rawPredict path
+	name       string
+	catalogID  string
+	upstream   string
+	apiKey     string
+	model      string // body model (chat/messages) or path model@version (vertex)
+	kind       string // harness.WireChat, harness.WireMessages, or harness.WireVertex
+	project    string // vertex only: GCP project for the rawPredict path
+	region     string // vertex only: GCP region for the rawPredict path
+	pathPrefix string // base-URL path prefix the agent carries (e.g. "/anthropic" for Kimi)
 }
 
 // availableProviders builds the matrix from the provider env vars that are set.
@@ -39,6 +248,24 @@ func availableProviders() []providerCase {
 	if k := os.Getenv("ANTHROPIC_TOKEN"); k != "" {
 		ps = append(ps, providerCase{name: "anthropic", catalogID: "anthropic_api", upstream: "https://api.anthropic.com", apiKey: k, model: "claude-haiku-4-5", kind: harness.WireMessages})
 	}
+	if k := os.Getenv("KIMI_TOKEN"); k != "" {
+		// Kimi (Moonshot AI) serves two body shapes from the same key: OpenAI
+		// Chat Completions on the bare host (/v1/...) and the Anthropic
+		// Messages API under the /anthropic path prefix (the endpoint
+		// Moonshot's Claude Code guide uses). The provider keeps the bare
+		// default upstream and the AGENT carries the /anthropic prefix in
+		// its base URL — exactly the documented Claude Code / Kimi CLI
+		// setup (ANTHROPIC_BASE_URL=https://<endpoint>/anthropic) — so one
+		// provider serves both shapes and the prefix rides through to
+		// Moonshot. Run the Anthropic shape, the flagship Claude Code path;
+		// the OpenAI wire shape is covered live by the other chat-shaped
+		// matrix providers, and Kimi-over-chat passed with kimi-k3 before
+		// the single-model constraint surfaced (run #73 on the kimi feature
+		// branch). The platform serves this account exactly ONE model —
+		// kimi-k3 (kimi-k2-thinking and even kimi-latest return
+		// resource_not_found_error on both surfaces).
+		ps = append(ps, providerCase{name: "kimi", catalogID: "kimi_api", upstream: "https://api.moonshot.ai", apiKey: k, model: "kimi-k3", kind: harness.WireMessages, pathPrefix: "/anthropic"})
+	}
 	if k, u := os.Getenv("VERCEL_TOKEN"), os.Getenv("VERCEL_URL"); k != "" && u != "" {
 		ps = append(ps, providerCase{name: "vercel", catalogID: "vercel_ai_gateway", upstream: u, apiKey: k, model: "openai/gpt-4o-mini", kind: harness.WireChat})
 	}
@@ -84,14 +311,27 @@ func availableProviders() []providerCase {
 		}
 	}
 
-	// Bedrock: path-routed, bearer auth. Model is a cross-region inference
-	// profile id (distinct string from the first-party Anthropic case).
+	// Bedrock: path-routed, bearer auth. Model is the FULL cross-region
+	// inference-profile id exactly as AWS issues it — region-family prefix
+	// plus the date/version suffix. A bare or wrong-region id makes Bedrock
+	// reject the request with "The provided model identifier is invalid"
+	// before any inference runs. The proxy normalizes this id to the catalog
+	// key (anthropic.claude-haiku-4-5) for routing/pricing/allowlists.
+	// Defaults pair eu-central-1 with the eu.* profile; AWS_REGION overrides
+	// the region and the prefix follows its family.
 	if k := os.Getenv("AWS_BEARER_TOKEN_BEDROCK"); k != "" {
 		region := os.Getenv("AWS_REGION")
 		if region == "" {
-			region = "us-east-1"
+			region = "eu-central-1"
 		}
-		ps = append(ps, providerCase{name: "bedrock", catalogID: "bedrock_api", upstream: "https://bedrock-runtime." + region + ".amazonaws.com", apiKey: k, model: "us.anthropic.claude-haiku-4-5", kind: harness.WireBedrock})
+		// A valid Bedrock inference-profile id, overridable per account (AWS_BEDROCK_MODEL, also the
+		// workflow's bedrock_model dispatch input). `global.` profiles work from any region. Defaults to
+		// Sonnet 4.6, whose id convention dropped the -YYYYMMDD-v1:0 suffix that Haiku 4.5 still carries.
+		model := os.Getenv("AWS_BEDROCK_MODEL")
+		if model == "" {
+			model = "global.anthropic.claude-sonnet-4-6"
+		}
+		ps = append(ps, providerCase{name: "bedrock", catalogID: "bedrock_api", upstream: "https://bedrock-runtime." + region + ".amazonaws.com", apiKey: k, model: model, kind: harness.WireBedrock})
 	}
 	return ps
 }
@@ -108,8 +348,16 @@ func providerRequest(pc providerCase) api.AgentNetworkProviderRequest {
 		Enabled:     ptr(true),
 	}
 	if pc.kind != harness.WireVertex {
+		// The router matches the normalized catalog id. Bedrock's request model
+		// travels as a region-prefixed inference-profile id in the URL path
+		// (us.anthropic...), which the router strips before matching, so register
+		// the normalized form here or routing fails as model_not_routable.
+		modelID := pc.model
+		if pc.kind == harness.WireBedrock {
+			modelID = catalogModel(pc)
+		}
 		req.Models = &[]api.AgentNetworkProviderModel{
-			{Id: pc.model, InputPer1k: 0.001, OutputPer1k: 0.002},
+			{Id: modelID, InputPer1k: 0.001, OutputPer1k: 0.002},
 		}
 	}
 	return req
@@ -201,11 +449,12 @@ func TestProvidersMatrix(t *testing.T) {
 	t.Cleanup(func() { _ = cl.Terminate(context.Background()) })
 
 	require.NoError(t, cl.WaitConnected(ctx, 90*time.Second), "client must connect to management")
+	// Probe first: the GET resolves the endpoint (DNS error fails) and its first packet wakes the lazy proxy peer, so WaitProxyPeer sees it connected; any HTTP status counts.
+	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
+	require.NoError(t, err, "resolve agent-network endpoint to proxy IP")
 	if err := cl.WaitProxyPeer(ctx, 180*time.Second); err != nil {
 		t.Fatalf("client did not see the proxy peer: %v\n=== proxy logs ===\n%s", err, px.Logs(context.Background()))
 	}
-	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
-	require.NoError(t, err, "resolve agent-network endpoint to proxy IP")
 
 	for _, pc := range matrix {
 		pc := pc
@@ -216,6 +465,10 @@ func TestProvidersMatrix(t *testing.T) {
 			// session id and confirm the marker propagated end-to-end.
 			sessionID := "e2e-session-" + pc.name
 
+			// A long-form prompt so completions carry realistic token counts for cost validation;
+			// max_tokens in the harness bodies (2048) lets the full answer through.
+			const matrixPrompt = "explain GitHub workflow in 1000 words"
+
 			// Retry briefly to absorb tunnel/DNS jitter on the first call.
 			var code int
 			var body string
@@ -226,11 +479,11 @@ func TestProvidersMatrix(t *testing.T) {
 				var cerr error
 				switch pc.kind {
 				case harness.WireVertex:
-					c, b, cerr = cl.Vertex(ctx, settings.Endpoint, proxyIP, pc.project, pc.region, pc.model, "Reply with exactly: pong", sessionID)
+					c, b, cerr = cl.Vertex(ctx, settings.Endpoint, proxyIP, pc.project, pc.region, pc.model, matrixPrompt, sessionID)
 				case harness.WireBedrock:
-					c, b, cerr = cl.Bedrock(ctx, settings.Endpoint, proxyIP, pc.model, "Reply with exactly: pong", sessionID)
+					c, b, cerr = cl.Bedrock(ctx, settings.Endpoint, proxyIP, pc.model, matrixPrompt, sessionID)
 				default:
-					c, b, cerr = cl.Chat(ctx, settings.Endpoint, proxyIP, pc.kind, pc.model, "Reply with exactly: pong", sessionID)
+					c, b, cerr = cl.ChatPrefixed(ctx, settings.Endpoint, proxyIP, pc.pathPrefix, pc.kind, pc.model, matrixPrompt, sessionID)
 				}
 				if cerr == nil {
 					code, body = c, b
@@ -249,6 +502,7 @@ func TestProvidersMatrix(t *testing.T) {
 
 			// The session id sent as x-session-id must round-trip into the
 			// access-log row for this provider.
+			var row api.AgentNetworkAccessLog
 			require.Eventually(t, func() bool {
 				logs, lerr := srv.ListAccessLogs(ctx)
 				if lerr != nil {
@@ -256,11 +510,15 @@ func TestProvidersMatrix(t *testing.T) {
 				}
 				for _, r := range logs.Data {
 					if r.SessionId != nil && *r.SessionId == sessionID {
+						row = r
 						return true
 					}
 				}
 				return false
 			}, 30*time.Second, 2*time.Second, "session id %q must be recorded in an access-log row for %s", sessionID, pc.name)
+
+			// Stored total and cache cost must match the published rates applied to the row's buckets.
+			validateAccessLogCost(t, pc, row)
 		})
 	}
 
@@ -281,4 +539,7 @@ func TestProvidersMatrix(t *testing.T) {
 		}
 		return false
 	}, 60*time.Second, 3*time.Second, "consumption must be recorded with positive token counts after live traffic")
+
+	// Final raw-SQL audit: bypass the API and re-verify every persisted usage row in the store.
+	verifyUsageRowsSQL(t, srv)
 }

e2e/agentnetwork/guardrail_block_test.go

@@ -0,0 +1,209 @@
+//go:build e2e
+
+package agentnetwork
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/e2e/harness"
+	"github.com/netbirdio/netbird/shared/management/http/api"
+)
+
+// pathRoutedGuardrailCase is one provider's self-contained scenario: its own
+// provider, its own guardrail whose allowlist holds ONLY that provider's
+// allowed model, and its own policy. Each case runs in isolation (its own
+// proxy + client), so the guardrail the proxy enforces contains exactly this
+// provider's model — never a mixed cross-provider list.
+type pathRoutedGuardrailCase struct {
+	name       string
+	catalogID  string // agent-network catalog provider id
+	wire       string // harness.WireVertex | harness.WireBedrock
+	allowEntry string // the single model id put on the guardrail allowlist
+	allowModel string // model id sent that MUST be served (200)
+	blockModel string // model id sent that MUST be denied (403 model_blocked)
+}
+
+// TestGuardrailBlocksUnselectedModel_PathRouted is the end-to-end regression
+// guard for the customer report that a model-allowlist guardrail attached to a
+// policy has no effect for PATH-ROUTED providers — where the model travels in
+// the URL, not the JSON body: Google Vertex (…/models/{model}:rawPredict) and
+// AWS Bedrock (/model/{id}/invoke).
+//
+// Each provider is tested in isolation with a guardrail allowlisting a single
+// model of its own: the allowed model (in the URL path) is served (200) and an
+// unselected model (in the URL path) is denied 403 by the guardrail
+// (llm_policy.model_blocked) before the upstream. The Vertex case mirrors the
+// customer verbatim — allow Sonnet, and the unselected model is the exact
+// claude-opus-4-6 they reported reaching the model unblocked. The Bedrock case
+// sends a region-prefixed, versioned inference-profile id so URL-path model
+// normalization is exercised too.
+//
+// The provider is catch-all (no models), so the router forwards any model and a
+// 403 can only come from the guardrail, never model_not_routable. Only the
+// upstream LLM is mocked (the vLLM nginx answers any path with 200); management
+// synth/reconcile, the proxy middleware chain (URL-path model extraction,
+// router, guardrail) and the tunnel are all real, and the guardrail denies
+// before the upstream is dialed so the mock cannot influence the block. A
+// static bearer api key is used so the router injects a static Authorization
+// header instead of minting a GCP token — the only reason path-routed providers
+// normally need live credentials — so the test runs with none and is always on.
+func TestGuardrailBlocksUnselectedModel_PathRouted(t *testing.T) {
+	cases := []pathRoutedGuardrailCase{
+		{
+			name:       "vertex",
+			catalogID:  "vertex_ai_api",
+			wire:       harness.WireVertex,
+			allowEntry: "claude-sonnet-4-5",
+			allowModel: "claude-sonnet-4-5",
+			blockModel: "claude-opus-4-6", // the customer-reported model
+		},
+		{
+			name:       "bedrock",
+			catalogID:  "bedrock_api",
+			wire:       harness.WireBedrock,
+			allowEntry: "anthropic.claude-sonnet-4-5", // normalized catalog id
+			allowModel: "us.anthropic.claude-sonnet-4-5-v1:0",
+			blockModel: "us.anthropic.claude-opus-4-8-v1:0",
+		},
+	}
+
+	for _, tc := range cases {
+		tc := tc
+		t.Run(tc.name, func(t *testing.T) {
+			runPathRoutedGuardrailCase(t, tc)
+		})
+	}
+}
+
+func runPathRoutedGuardrailCase(t *testing.T, tc pathRoutedGuardrailCase) {
+	t.Helper()
+
+	const (
+		vertexProject = "e2e-project"
+		vertexRegion  = "global"
+	)
+
+	ctx, cancel := context.WithTimeout(context.Background(), 15*time.Minute)
+	defer cancel()
+
+	vllm, err := harness.StartVLLM(ctx, srv)
+	require.NoError(t, err, "start mock upstream")
+	t.Cleanup(func() { _ = vllm.Terminate(context.Background()) })
+
+	grp, err := srv.API().Groups.Create(ctx, api.PostApiGroupsJSONRequestBody{Name: "e2e-guardrail-" + tc.name})
+	require.NoError(t, err, "create group")
+	t.Cleanup(func() { _ = srv.API().Groups.Delete(context.Background(), grp.Id) })
+
+	ephemeral := false
+	sk, err := srv.API().SetupKeys.Create(ctx, api.PostApiSetupKeysJSONRequestBody{
+		Name:       "e2e-guardrail-" + tc.name + "-client",
+		Type:       "reusable",
+		ExpiresIn:  86400,
+		UsageLimit: 0,
+		AutoGroups: []string{grp.Id},
+		Ephemeral:  &ephemeral,
+	})
+	require.NoError(t, err, "mint setup key")
+	require.NotEmpty(t, sk.Key, "setup key plaintext")
+
+	// Catch-all provider (no models) so the router forwards any model; a static
+	// bearer key means the router injects a static auth header instead of minting
+	// a GCP token. Bootstraps the cluster if it isn't already.
+	staticKey := "static-e2e-token"
+	prov, err := srv.CreateProvider(ctx, api.AgentNetworkProviderRequest{
+		Name:             tc.name,
+		ProviderId:       tc.catalogID,
+		UpstreamUrl:      vllm.URL,
+		ApiKey:           &staticKey,
+		Enabled:          ptr(true),
+		BootstrapCluster: ptr(harness.AgentNetworkCluster),
+	})
+	require.NoError(t, err, "create %s provider", tc.name)
+	t.Cleanup(func() { _ = srv.DeleteProvider(context.Background(), prov.Id) })
+
+	// Guardrail allowlisting ONLY this provider's allowed model.
+	var gr api.AgentNetworkGuardrailRequest
+	gr.Name = "e2e-guardrail-" + tc.name
+	gr.Checks.ModelAllowlist.Enabled = true
+	gr.Checks.ModelAllowlist.Models = []string{tc.allowEntry}
+	guard, err := srv.CreateGuardrail(ctx, gr)
+	require.NoError(t, err, "create guardrail")
+	t.Cleanup(func() { _ = srv.DeleteGuardrail(context.Background(), guard.Id) })
+
+	enabled := true
+	pol, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-guardrail-" + tc.name,
+		Enabled:                &enabled,
+		SourceGroups:           []string{grp.Id},
+		DestinationProviderIds: []string{prov.Id},
+		GuardrailIds:           &[]string{guard.Id},
+	})
+	require.NoError(t, err, "create policy")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), pol.Id) })
+
+	settings, err := srv.GetSettings(ctx)
+	require.NoError(t, err, "read settings")
+	require.NotEmpty(t, settings.Endpoint, "endpoint must be assigned")
+
+	proxyToken, err := srv.CreateProxyTokenCLI(ctx, "e2e-guardrail-"+tc.name+"-proxy")
+	require.NoError(t, err, "mint proxy token")
+	px, err := harness.StartProxy(ctx, srv, proxyToken)
+	require.NoError(t, err, "start proxy")
+	t.Cleanup(func() { _ = px.Terminate(context.Background()) })
+
+	cl, err := harness.StartClient(ctx, srv, sk.Key)
+	require.NoError(t, err, "start client")
+	t.Cleanup(func() { _ = cl.Terminate(context.Background()) })
+
+	require.NoError(t, cl.WaitConnected(ctx, 90*time.Second), "client must connect to management")
+	// Probe first: the GET resolves the endpoint and its first packet wakes the
+	// lazy proxy peer, so WaitProxyPeer then observes it connected.
+	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
+	require.NoError(t, err, "resolve endpoint to proxy IP")
+	if err := cl.WaitProxyPeer(ctx, 180*time.Second); err != nil {
+		t.Fatalf("client did not see the proxy peer: %v\n=== proxy logs ===\n%s", err, px.Logs(context.Background()))
+	}
+
+	send := func(model string) (int, string) {
+		var code int
+		var body string
+		var cerr error
+		switch tc.wire {
+		case harness.WireVertex:
+			code, body, cerr = cl.Vertex(ctx, settings.Endpoint, proxyIP, vertexProject, vertexRegion, model, "Reply with exactly: pong", "")
+		case harness.WireBedrock:
+			code, body, cerr = cl.Bedrock(ctx, settings.Endpoint, proxyIP, model, "Reply with exactly: pong", "")
+		default:
+			t.Fatalf("unsupported wire %q", tc.wire)
+		}
+		require.NoError(t, cerr, "request must reach the proxy for %s", tc.name)
+		return code, body
+	}
+
+	// Allowed model (in the URL path) is served. Retry to absorb tunnel/DNS
+	// jitter on the first call over the freshly warmed tunnel.
+	var code int
+	var body string
+	deadline := time.Now().Add(90 * time.Second)
+	for time.Now().Before(deadline) {
+		code, body = send(tc.allowModel)
+		if code == 200 {
+			break
+		}
+		time.Sleep(5 * time.Second)
+	}
+	assert.Equal(t, 200, code,
+		"allowed %s model (URL path) must be served; body: %s\n=== proxy logs ===\n%s", tc.name, body, px.Logs(context.Background()))
+
+	// Unselected model (in the URL path) must be blocked by the guardrail.
+	code, body = send(tc.blockModel)
+	assert.Equal(t, 403, code,
+		"unselected %s model (URL path) must be denied, not served; body: %s\n=== proxy logs ===\n%s", tc.name, body, px.Logs(context.Background()))
+	assert.Contains(t, body, "llm_policy.model_blocked",
+		"%s denial must come from the guardrail allowlist, not routing; body: %s", tc.name, body)
+}

e2e/agentnetwork/guardrail_groupswitch_test.go

@@ -0,0 +1,205 @@
+//go:build e2e
+
+package agentnetwork
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/e2e/harness"
+	"github.com/netbirdio/netbird/shared/management/http/api"
+)
+
+// TestGuardrailGroupSwitchTakesEffectAfterTTL proves that moving a peer between
+// groups flips its model-allowlist decision once the proxy's tunnel-peer cache
+// expires. The peer's groups reach the guardrail via ValidateTunnelPeer, which
+// the proxy caches; the switch is invisible until that cache expires. The proxy
+// runs with a short NB_PROXY_TUNNEL_CACHE_TTL so the flip happens in seconds
+// instead of the 5-minute default.
+//
+// Setup: one catch-all provider declaring modelA + modelB; polA (grpA -> allow
+// modelA) and polB (grpB -> allow modelB). The client starts in grpA. modelA is
+// served and modelB denied; after switching the client grpA -> grpB, modelB is
+// served and modelA denied. The cross-group deny comes from management's
+// per-policy/group CheckLLMPolicyLimits (the proxy backstop carries the union).
+func TestGuardrailGroupSwitchTakesEffectAfterTTL(t *testing.T) {
+	ctx, cancel := context.WithTimeout(context.Background(), 15*time.Minute)
+	defer cancel()
+
+	const (
+		modelA = "e2e-model-a"
+		modelB = "e2e-model-b"
+		// Short tunnel-cache TTL so a group switch propagates in seconds.
+		// Exercises the NB_PROXY_TUNNEL_CACHE_TTL override.
+		cacheTTL = 3 * time.Second
+	)
+
+	vllm, err := harness.StartVLLM(ctx, srv)
+	require.NoError(t, err, "start mock upstream")
+	t.Cleanup(func() { _ = vllm.Terminate(context.Background()) })
+
+	grpA, err := srv.API().Groups.Create(ctx, api.PostApiGroupsJSONRequestBody{Name: "e2e-gswitch-a"})
+	require.NoError(t, err, "create group A")
+	t.Cleanup(func() { _ = srv.API().Groups.Delete(context.Background(), grpA.Id) })
+
+	grpB, err := srv.API().Groups.Create(ctx, api.PostApiGroupsJSONRequestBody{Name: "e2e-gswitch-b"})
+	require.NoError(t, err, "create group B")
+	t.Cleanup(func() { _ = srv.API().Groups.Delete(context.Background(), grpB.Id) })
+
+	ephemeral := false
+	sk, err := srv.API().SetupKeys.Create(ctx, api.PostApiSetupKeysJSONRequestBody{
+		Name:       "e2e-gswitch-client",
+		Type:       "reusable",
+		ExpiresIn:  86400,
+		UsageLimit: 0,
+		AutoGroups: []string{grpA.Id}, // client starts in group A
+		Ephemeral:  &ephemeral,
+	})
+	require.NoError(t, err, "mint setup key")
+	require.NotEmpty(t, sk.Key, "setup key plaintext")
+
+	staticKey := "static-e2e-token"
+	prov, err := srv.CreateProvider(ctx, api.AgentNetworkProviderRequest{
+		Name:        "gswitch",
+		ProviderId:  "openai_api",
+		UpstreamUrl: vllm.URL,
+		ApiKey:      &staticKey,
+		Enabled:     ptr(true),
+		Models: &[]api.AgentNetworkProviderModel{
+			{Id: modelA, InputPer1k: 0.001, OutputPer1k: 0.001},
+			{Id: modelB, InputPer1k: 0.001, OutputPer1k: 0.001},
+		},
+		BootstrapCluster: ptr(harness.AgentNetworkCluster),
+	})
+	require.NoError(t, err, "create provider")
+	t.Cleanup(func() { _ = srv.DeleteProvider(context.Background(), prov.Id) })
+
+	mkGuard := func(name, model string) api.AgentNetworkGuardrail {
+		var gr api.AgentNetworkGuardrailRequest
+		gr.Name = name
+		gr.Checks.ModelAllowlist.Enabled = true
+		gr.Checks.ModelAllowlist.Models = []string{model}
+		g, gerr := srv.CreateGuardrail(ctx, gr)
+		require.NoError(t, gerr, "create guardrail %s", name)
+		t.Cleanup(func() { _ = srv.DeleteGuardrail(context.Background(), g.Id) })
+		return g
+	}
+	gA := mkGuard("e2e-gswitch-a", modelA)
+	gB := mkGuard("e2e-gswitch-b", modelB)
+
+	enabled := true
+	polA, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-gswitch-a",
+		Enabled:                &enabled,
+		SourceGroups:           []string{grpA.Id},
+		DestinationProviderIds: []string{prov.Id},
+		GuardrailIds:           &[]string{gA.Id},
+	})
+	require.NoError(t, err, "create policy A")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polA.Id) })
+
+	polB, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-gswitch-b",
+		Enabled:                &enabled,
+		SourceGroups:           []string{grpB.Id},
+		DestinationProviderIds: []string{prov.Id},
+		GuardrailIds:           &[]string{gB.Id},
+	})
+	require.NoError(t, err, "create policy B")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polB.Id) })
+
+	settings, err := srv.GetSettings(ctx)
+	require.NoError(t, err, "read settings")
+	require.NotEmpty(t, settings.Endpoint, "endpoint must be assigned")
+
+	proxyToken, err := srv.CreateProxyTokenCLI(ctx, "e2e-gswitch-proxy")
+	require.NoError(t, err, "mint proxy token")
+	px, err := harness.StartProxy(ctx, srv, proxyToken, map[string]string{
+		"NB_PROXY_TUNNEL_CACHE_TTL": cacheTTL.String(),
+	})
+	require.NoError(t, err, "start proxy")
+	t.Cleanup(func() { _ = px.Terminate(context.Background()) })
+
+	cl, err := harness.StartClient(ctx, srv, sk.Key)
+	require.NoError(t, err, "start client")
+	t.Cleanup(func() { _ = cl.Terminate(context.Background()) })
+
+	require.NoError(t, cl.WaitConnected(ctx, 90*time.Second), "client must connect to management")
+	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
+	require.NoError(t, err, "resolve endpoint to proxy IP")
+	if err := cl.WaitProxyPeer(ctx, 180*time.Second); err != nil {
+		t.Fatalf("client did not see the proxy peer: %v\n=== proxy logs ===\n%s", err, px.Logs(context.Background()))
+	}
+
+	send := func(model string) (int, string) {
+		code, body, cerr := cl.Chat(ctx, settings.Endpoint, proxyIP, harness.WireChat, model, "Reply with exactly: pong", "")
+		require.NoError(t, cerr, "request must reach the proxy")
+		return code, body
+	}
+	sendUntil := func(model string, want int, timeout time.Duration) (int, string) {
+		var code int
+		var body string
+		deadline := time.Now().Add(timeout)
+		for time.Now().Before(deadline) {
+			code, body = send(model)
+			if code == want {
+				return code, body
+			}
+			time.Sleep(2 * time.Second)
+		}
+		return code, body
+	}
+
+	// Phase 1 — client is in group A: modelA served, modelB denied.
+	code, body := sendUntil(modelA, 200, 90*time.Second)
+	assert.Equal(t, 200, code,
+		"group-A model must be served while the client is in group A; body: %s\n=== proxy logs ===\n%s", body, px.Logs(context.Background()))
+	code, body = send(modelB)
+	assert.Equal(t, 403, code,
+		"group-B model must be denied while the client is in group A; body: %s", body)
+	assert.Contains(t, body, "llm_policy.model_blocked")
+
+	// Switch the client peer from group A to group B.
+	peerID := clientPeerInGroup(t, ctx, grpA.Id)
+	_, err = srv.API().Groups.Update(ctx, grpB.Id, api.PutApiGroupsGroupIdJSONRequestBody{
+		Name:  grpB.Name,
+		Peers: &[]string{peerID},
+	})
+	require.NoError(t, err, "add peer to group B")
+	_, err = srv.API().Groups.Update(ctx, grpA.Id, api.PutApiGroupsGroupIdJSONRequestBody{
+		Name:  grpA.Name,
+		Peers: &[]string{},
+	})
+	require.NoError(t, err, "remove peer from group A")
+
+	// Phase 2 — after the short TTL expires the proxy re-validates the peer,
+	// sees group B, and the decision flips. Poll to absorb TTL + re-validation.
+	code, body = sendUntil(modelB, 200, 60*time.Second)
+	assert.Equal(t, 200, code,
+		"after the group switch + TTL, the group-B model must be served; body: %s\n=== proxy logs ===\n%s", body, px.Logs(context.Background()))
+	code, body = send(modelA)
+	assert.Equal(t, 403, code,
+		"after the switch, the old group-A model must be denied; body: %s", body)
+	assert.Contains(t, body, "llm_policy.model_blocked")
+}
+
+// clientPeerInGroup returns the id of the single peer that is a member of the
+// given group — the test client. The proxy peer is never added to test groups.
+func clientPeerInGroup(t *testing.T, ctx context.Context, groupID string) string {
+	t.Helper()
+	peers, err := srv.API().Peers.List(ctx)
+	require.NoError(t, err, "list peers")
+	for _, p := range peers {
+		for _, g := range p.Groups {
+			if g.Id == groupID {
+				return p.Id
+			}
+		}
+	}
+	t.Fatalf("no peer found in group %s", groupID)
+	return ""
+}

e2e/agentnetwork/guardrail_multipolicy_test.go

@@ -0,0 +1,201 @@
+//go:build e2e
+
+package agentnetwork
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/e2e/harness"
+	"github.com/netbirdio/netbird/shared/management/http/api"
+)
+
+// TestGuardrailMultiPolicyModelAllowlist: modelSelected served (200), grpOther's
+// modelOther denied for the grpMain client (403 model_blocked, no cross-group
+// leak), and openModel on the un-guardrailed policy's provider served (200).
+func TestGuardrailMultiPolicyModelAllowlist(t *testing.T) {
+	ctx, cancel := context.WithTimeout(context.Background(), 15*time.Minute)
+	defer cancel()
+
+	const (
+		modelSelected = "e2e-selected"
+		modelOther    = "e2e-other"
+		openModel     = "e2e-open"
+	)
+
+	vllm, err := harness.StartVLLM(ctx, srv)
+	require.NoError(t, err, "start mock upstream")
+	t.Cleanup(func() { _ = vllm.Terminate(context.Background()) })
+
+	grpMain, err := srv.API().Groups.Create(ctx, api.PostApiGroupsJSONRequestBody{Name: "e2e-guardrail-mp-main"})
+	require.NoError(t, err, "create main group")
+	t.Cleanup(func() { _ = srv.API().Groups.Delete(context.Background(), grpMain.Id) })
+
+	grpOther, err := srv.API().Groups.Create(ctx, api.PostApiGroupsJSONRequestBody{Name: "e2e-guardrail-mp-other"})
+	require.NoError(t, err, "create other group")
+	t.Cleanup(func() { _ = srv.API().Groups.Delete(context.Background(), grpOther.Id) })
+
+	ephemeral := false
+	sk, err := srv.API().SetupKeys.Create(ctx, api.PostApiSetupKeysJSONRequestBody{
+		Name:       "e2e-guardrail-mp-client",
+		Type:       "reusable",
+		ExpiresIn:  86400,
+		UsageLimit: 0,
+		AutoGroups: []string{grpMain.Id}, // client joins grpMain only
+		Ephemeral:  &ephemeral,
+	})
+	require.NoError(t, err, "mint setup key")
+	require.NotEmpty(t, sk.Key, "setup key plaintext")
+
+	staticKey := "static-e2e-token"
+	models := func(ids ...string) *[]api.AgentNetworkProviderModel {
+		out := make([]api.AgentNetworkProviderModel, 0, len(ids))
+		for _, id := range ids {
+			out = append(out, api.AgentNetworkProviderModel{Id: id, InputPer1k: 0.001, OutputPer1k: 0.001})
+		}
+		return &out
+	}
+
+	// pRestricted declares the two guardrailed models so routing is deterministic
+	// (model -> provider). Created first, so it carries the bootstrap cluster.
+	pRestricted, err := srv.CreateProvider(ctx, api.AgentNetworkProviderRequest{
+		Name:             "restricted",
+		ProviderId:       "openai_api",
+		UpstreamUrl:      vllm.URL,
+		ApiKey:           &staticKey,
+		Enabled:          ptr(true),
+		Models:           models(modelSelected, modelOther),
+		BootstrapCluster: ptr(harness.AgentNetworkCluster),
+	})
+	require.NoError(t, err, "create restricted provider")
+	t.Cleanup(func() { _ = srv.DeleteProvider(context.Background(), pRestricted.Id) })
+
+	pOpen, err := srv.CreateProvider(ctx, api.AgentNetworkProviderRequest{
+		Name:        "open",
+		ProviderId:  "openai_api",
+		UpstreamUrl: vllm.URL,
+		ApiKey:      &staticKey,
+		Enabled:     ptr(true),
+		Models:      models(openModel),
+	})
+	require.NoError(t, err, "create open provider")
+	t.Cleanup(func() { _ = srv.DeleteProvider(context.Background(), pOpen.Id) })
+
+	mkGuardrail := func(name, model string) api.AgentNetworkGuardrail {
+		var gr api.AgentNetworkGuardrailRequest
+		gr.Name = name
+		gr.Checks.ModelAllowlist.Enabled = true
+		gr.Checks.ModelAllowlist.Models = []string{model}
+		g, gerr := srv.CreateGuardrail(ctx, gr)
+		require.NoError(t, gerr, "create guardrail %s", name)
+		t.Cleanup(func() { _ = srv.DeleteGuardrail(context.Background(), g.Id) })
+		return g
+	}
+	gMain := mkGuardrail("e2e-guardrail-mp-main", modelSelected)
+	gOther := mkGuardrail("e2e-guardrail-mp-other", modelOther)
+
+	enabled := true
+	// polMain: grpMain restricted to modelSelected on pRestricted.
+	polMain, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-guardrail-mp-main",
+		Enabled:                &enabled,
+		SourceGroups:           []string{grpMain.Id},
+		DestinationProviderIds: []string{pRestricted.Id},
+		GuardrailIds:           &[]string{gMain.Id},
+	})
+	require.NoError(t, err, "create main policy")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polMain.Id) })
+
+	// polOther: grpOther restricted to modelOther on the SAME provider. The
+	// client is not in grpOther, so modelOther must never be usable by it.
+	polOther, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-guardrail-mp-other",
+		Enabled:                &enabled,
+		SourceGroups:           []string{grpOther.Id},
+		DestinationProviderIds: []string{pRestricted.Id},
+		GuardrailIds:           &[]string{gOther.Id},
+	})
+	require.NoError(t, err, "create other policy")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polOther.Id) })
+
+	// polOpen: grpMain on pOpen with NO guardrail — unrestricted.
+	polOpen, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-guardrail-mp-open",
+		Enabled:                &enabled,
+		SourceGroups:           []string{grpMain.Id},
+		DestinationProviderIds: []string{pOpen.Id},
+	})
+	require.NoError(t, err, "create open policy")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polOpen.Id) })
+
+	settings, err := srv.GetSettings(ctx)
+	require.NoError(t, err, "read settings")
+	require.NotEmpty(t, settings.Endpoint, "endpoint must be assigned")
+
+	proxyToken, err := srv.CreateProxyTokenCLI(ctx, "e2e-guardrail-mp-proxy")
+	require.NoError(t, err, "mint proxy token")
+	px, err := harness.StartProxy(ctx, srv, proxyToken)
+	require.NoError(t, err, "start proxy")
+	t.Cleanup(func() { _ = px.Terminate(context.Background()) })
+
+	cl, err := harness.StartClient(ctx, srv, sk.Key)
+	require.NoError(t, err, "start client")
+	t.Cleanup(func() { _ = cl.Terminate(context.Background()) })
+
+	require.NoError(t, cl.WaitConnected(ctx, 90*time.Second), "client must connect to management")
+	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
+	require.NoError(t, err, "resolve endpoint to proxy IP")
+	if err := cl.WaitProxyPeer(ctx, 180*time.Second); err != nil {
+		t.Fatalf("client did not see the proxy peer: %v\n=== proxy logs ===\n%s", err, px.Logs(context.Background()))
+	}
+
+	send := func(model string) (int, string) {
+		code, body, cerr := cl.Chat(ctx, settings.Endpoint, proxyIP, harness.WireChat, model, "Reply with exactly: pong", "")
+		require.NoError(t, cerr, "request must reach the proxy")
+		return code, body
+	}
+	// sendUntil200 absorbs first-call tunnel/DNS jitter on the freshly warmed tunnel.
+	sendUntil200 := func(model string) (int, string) {
+		var code int
+		var body string
+		deadline := time.Now().Add(90 * time.Second)
+		for time.Now().Before(deadline) {
+			code, body = send(model)
+			if code == 200 {
+				break
+			}
+			time.Sleep(5 * time.Second)
+		}
+		return code, body
+	}
+
+	t.Run("selected model allowed for its group", func(t *testing.T) {
+		code, body := sendUntil200(modelSelected)
+		assert.Equal(t, 200, code,
+			"grpMain's allowlisted model must be served; body: %s\n=== proxy logs ===\n%s", body, px.Logs(context.Background()))
+	})
+
+	t.Run("other group's model does not leak", func(t *testing.T) {
+		// modelOther is allowlisted only for grpOther. The grpMain client must be
+		// denied by management's per-policy/group check — not waved through by an
+		// account-wide union. This is the security-critical wrong-ALLOW guard.
+		code, body := send(modelOther)
+		assert.Equal(t, 403, code,
+			"another group's allowlisted model must be denied for this caller; body: %s\n=== proxy logs ===\n%s", body, px.Logs(context.Background()))
+		assert.Contains(t, body, "llm_policy.model_blocked",
+			"denial must be a model-allowlist decision; body: %s", body)
+	})
+
+	t.Run("unguarded policy leaves its provider unrestricted", func(t *testing.T) {
+		// polOpen carries no guardrail, so pOpen is unrestricted for grpMain. The
+		// old account-wide union would have blocked openModel (it is on no
+		// allowlist); it must now be served — the false-DENY guard.
+		code, body := sendUntil200(openModel)
+		assert.Equal(t, 200, code,
+			"an un-guardrailed policy's provider must not be blocked by another policy's allowlist; body: %s\n=== proxy logs ===\n%s", body, px.Logs(context.Background()))
+	})
+}

e2e/agentnetwork/guardrail_pergroup_providers_test.go

@@ -0,0 +1,422 @@
+//go:build e2e
+
+package agentnetwork
+
+import (
+	"context"
+	"testing"
+	"time"
+
+	"github.com/stretchr/testify/assert"
+	"github.com/stretchr/testify/require"
+
+	"github.com/netbirdio/netbird/e2e/harness"
+	"github.com/netbirdio/netbird/shared/management/http/api"
+)
+
+// pergroupCase describes one provider surface for the per-group allowlist matrix.
+// selectedReq/otherReq are the model identifiers as they travel in the request
+// (URL path for Bedrock/Vertex, body "model" for chat/messages). selectedAllow/
+// otherAllow are the (normalized) forms the guardrail allowlist holds — for
+// Bedrock these differ from the request form so path normalization is exercised.
+type pergroupCase struct {
+	name          string
+	catalogID     string
+	wire          string // "chat", "messages", "vertex", "bedrock"
+	models        *[]api.AgentNetworkProviderModel
+	selectedReq   string
+	selectedAllow string
+	otherReq      string
+	otherAllow    string
+
+	providerID string // filled during setup
+}
+
+// TestGuardrailPerGroupAllowlist_AllProviders proves the per-policy/group model
+// allowlist end to end across every always-on provider surface, including the
+// path-routed ones (Vertex, Bedrock) where the model travels in the URL.
+//
+// For each provider two policies target it: grpMain (the client) is allowed only
+// selectedReq; grpOther (which the client is NOT in) is allowed only otherReq.
+// The client must get selectedReq served (200) and otherReq denied (403,
+// llm_policy.model_blocked) — the cross-group no-leak property. The deny is the
+// authoritative per-policy/group decision from management (the proxy per-provider
+// backstop carries the union of both models), so this also confirms management
+// receives the correct normalized model for path-routed providers.
+func TestGuardrailPerGroupAllowlist_AllProviders(t *testing.T) {
+	ctx, cancel := context.WithTimeout(context.Background(), 20*time.Minute)
+	defer cancel()
+
+	const (
+		vertexProject = "e2e-project"
+		vertexRegion  = "global"
+	)
+
+	priced := func(ids ...string) *[]api.AgentNetworkProviderModel {
+		out := make([]api.AgentNetworkProviderModel, 0, len(ids))
+		for _, id := range ids {
+			out = append(out, api.AgentNetworkProviderModel{Id: id, InputPer1k: 0.001, OutputPer1k: 0.001})
+		}
+		return &out
+	}
+
+	cases := []*pergroupCase{
+		{
+			name: "openai", catalogID: "openai_api", wire: harness.WireChat,
+			models:      priced("oai-model-a", "oai-model-b"),
+			selectedReq: "oai-model-a", selectedAllow: "oai-model-a",
+			otherReq: "oai-model-b", otherAllow: "oai-model-b",
+		},
+		{
+			name: "anthropic", catalogID: "anthropic_api", wire: harness.WireMessages,
+			models:      priced("ant-model-a", "ant-model-b"),
+			selectedReq: "ant-model-a", selectedAllow: "ant-model-a",
+			otherReq: "ant-model-b", otherAllow: "ant-model-b",
+		},
+		{
+			// Vertex catalog ids travel bare in the rawPredict path.
+			name: "vertex", catalogID: "vertex_ai_api", wire: "vertex",
+			selectedReq: "claude-sonnet-4-5", selectedAllow: "claude-sonnet-4-5",
+			otherReq: "claude-opus-4-6", otherAllow: "claude-opus-4-6",
+		},
+		{
+			// Bedrock request ids are region-prefixed/versioned; the parser
+			// normalizes them to the catalog key the allowlist holds.
+			name: "bedrock", catalogID: "bedrock_api", wire: "bedrock",
+			selectedReq: "us.anthropic.claude-sonnet-4-5-v1:0", selectedAllow: "anthropic.claude-sonnet-4-5",
+			otherReq: "us.anthropic.claude-opus-4-8-v1:0", otherAllow: "anthropic.claude-opus-4-8",
+		},
+	}
+
+	vllm, err := harness.StartVLLM(ctx, srv)
+	require.NoError(t, err, "start mock upstream")
+	t.Cleanup(func() { _ = vllm.Terminate(context.Background()) })
+
+	grpMain, err := srv.API().Groups.Create(ctx, api.PostApiGroupsJSONRequestBody{Name: "e2e-pergroup-main"})
+	require.NoError(t, err, "create main group")
+	t.Cleanup(func() { _ = srv.API().Groups.Delete(context.Background(), grpMain.Id) })
+
+	grpOther, err := srv.API().Groups.Create(ctx, api.PostApiGroupsJSONRequestBody{Name: "e2e-pergroup-other"})
+	require.NoError(t, err, "create other group")
+	t.Cleanup(func() { _ = srv.API().Groups.Delete(context.Background(), grpOther.Id) })
+
+	ephemeral := false
+	sk, err := srv.API().SetupKeys.Create(ctx, api.PostApiSetupKeysJSONRequestBody{
+		Name:       "e2e-pergroup-client",
+		Type:       "reusable",
+		ExpiresIn:  86400,
+		UsageLimit: 0,
+		AutoGroups: []string{grpMain.Id},
+		Ephemeral:  &ephemeral,
+	})
+	require.NoError(t, err, "mint setup key")
+	require.NotEmpty(t, sk.Key, "setup key plaintext")
+
+	staticKey := "static-e2e-token"
+	enabled := true
+
+	for i, c := range cases {
+		req := api.AgentNetworkProviderRequest{
+			Name:        "e2e-pergroup-" + c.name,
+			ProviderId:  c.catalogID,
+			UpstreamUrl: vllm.URL,
+			ApiKey:      &staticKey,
+			Enabled:     ptr(true),
+			Models:      c.models,
+		}
+		if i == 0 {
+			req.BootstrapCluster = ptr(harness.AgentNetworkCluster)
+		}
+		prov, perr := srv.CreateProvider(ctx, req)
+		require.NoError(t, perr, "create provider %s", c.name)
+		c.providerID = prov.Id
+		t.Cleanup(func() { _ = srv.DeleteProvider(context.Background(), prov.Id) })
+
+		gSel := mkAllowGuardrail(t, ctx, "e2e-pergroup-"+c.name+"-sel", c.selectedAllow)
+		gOth := mkAllowGuardrail(t, ctx, "e2e-pergroup-"+c.name+"-oth", c.otherAllow)
+
+		polMain, merr := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+			Name:                   "e2e-pergroup-" + c.name + "-main",
+			Enabled:                &enabled,
+			SourceGroups:           []string{grpMain.Id},
+			DestinationProviderIds: []string{prov.Id},
+			GuardrailIds:           &[]string{gSel.Id},
+		})
+		require.NoError(t, merr, "create main policy %s", c.name)
+		t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polMain.Id) })
+
+		polOther, oerr := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+			Name:                   "e2e-pergroup-" + c.name + "-other",
+			Enabled:                &enabled,
+			SourceGroups:           []string{grpOther.Id},
+			DestinationProviderIds: []string{prov.Id},
+			GuardrailIds:           &[]string{gOth.Id},
+		})
+		require.NoError(t, oerr, "create other policy %s", c.name)
+		t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polOther.Id) })
+	}
+
+	settings, err := srv.GetSettings(ctx)
+	require.NoError(t, err, "read settings")
+	require.NotEmpty(t, settings.Endpoint, "endpoint must be assigned")
+
+	proxyToken, err := srv.CreateProxyTokenCLI(ctx, "e2e-pergroup-proxy")
+	require.NoError(t, err, "mint proxy token")
+	px, err := harness.StartProxy(ctx, srv, proxyToken)
+	require.NoError(t, err, "start proxy")
+	t.Cleanup(func() { _ = px.Terminate(context.Background()) })
+
+	cl, err := harness.StartClient(ctx, srv, sk.Key)
+	require.NoError(t, err, "start client")
+	t.Cleanup(func() { _ = cl.Terminate(context.Background()) })
+
+	require.NoError(t, cl.WaitConnected(ctx, 90*time.Second), "client must connect to management")
+	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
+	require.NoError(t, err, "resolve endpoint to proxy IP")
+	if err := cl.WaitProxyPeer(ctx, 180*time.Second); err != nil {
+		t.Fatalf("client did not see the proxy peer: %v\n=== proxy logs ===\n%s", err, px.Logs(context.Background()))
+	}
+
+	send := func(c *pergroupCase, model string) (int, string) {
+		var code int
+		var body string
+		var cerr error
+		switch c.wire {
+		case "vertex":
+			code, body, cerr = cl.Vertex(ctx, settings.Endpoint, proxyIP, vertexProject, vertexRegion, model, "Reply with exactly: pong", "")
+		case "bedrock":
+			code, body, cerr = cl.Bedrock(ctx, settings.Endpoint, proxyIP, model, "Reply with exactly: pong", "")
+		default:
+			code, body, cerr = cl.Chat(ctx, settings.Endpoint, proxyIP, c.wire, model, "Reply with exactly: pong", "")
+		}
+		require.NoError(t, cerr, "request must reach the proxy for %s", c.name)
+		return code, body
+	}
+
+	for _, c := range cases {
+		t.Run(c.name, func(t *testing.T) {
+			// grpMain's own model is served. Retry to absorb tunnel/DNS jitter on
+			// the first call over the freshly warmed tunnel.
+			var code int
+			var body string
+			deadline := time.Now().Add(90 * time.Second)
+			for time.Now().Before(deadline) {
+				code, body = send(c, c.selectedReq)
+				if code == 200 {
+					break
+				}
+				time.Sleep(5 * time.Second)
+			}
+			assert.Equal(t, 200, code,
+				"%s: grpMain's allowlisted model must be served; body: %s\n=== proxy logs ===\n%s", c.name, body, px.Logs(context.Background()))
+
+			// grpOther's model must NOT leak to the grpMain client.
+			code, body = send(c, c.otherReq)
+			assert.Equal(t, 403, code,
+				"%s: another group's allowlisted model must be denied for this caller; body: %s\n=== proxy logs ===\n%s", c.name, body, px.Logs(context.Background()))
+			assert.Contains(t, body, "llm_policy.model_blocked",
+				"%s: denial must be a model-allowlist decision, not routing; body: %s", c.name, body)
+		})
+	}
+}
+
+// TestGuardrailMultiGroupUser proves the per-policy/group decision for a caller
+// that belongs to MULTIPLE groups at once. Two scenarios, one shared stack:
+//
+//   - union across the user's groups: the client is in gUX and gUY, each with
+//     its own policy+guardrail on provider P1 (gUX->union-a, gUY->union-b). The
+//     client may use BOTH models (the union of its groups' allowlists) while a
+//     third, un-allowlisted model is denied.
+//   - an un-guardrailed group lifts the restriction: the client is in gMP and
+//     gMQ on provider P2, where gMP restricts to mix-a but gMQ's policy carries
+//     NO guardrail. Because one applicable policy is unrestricted, the client may
+//     use a model on no allowlist (mix-z) as well as mix-a.
+func TestGuardrailMultiGroupUser(t *testing.T) {
+	ctx, cancel := context.WithTimeout(context.Background(), 15*time.Minute)
+	defer cancel()
+
+	const (
+		unionA = "mg-union-a"
+		unionB = "mg-union-b"
+		unionC = "mg-union-c" // allowlisted by neither group
+		mixA   = "mg-mix-a"
+		mixZ   = "mg-mix-z" // on no allowlist; reachable only via the un-guardrailed policy
+	)
+
+	priced := func(ids ...string) *[]api.AgentNetworkProviderModel {
+		out := make([]api.AgentNetworkProviderModel, 0, len(ids))
+		for _, id := range ids {
+			out = append(out, api.AgentNetworkProviderModel{Id: id, InputPer1k: 0.001, OutputPer1k: 0.001})
+		}
+		return &out
+	}
+
+	vllm, err := harness.StartVLLM(ctx, srv)
+	require.NoError(t, err, "start mock upstream")
+	t.Cleanup(func() { _ = vllm.Terminate(context.Background()) })
+
+	mkGroup := func(name string) *api.Group {
+		g, gerr := srv.API().Groups.Create(ctx, api.PostApiGroupsJSONRequestBody{Name: name})
+		require.NoError(t, gerr, "create group %s", name)
+		t.Cleanup(func() { _ = srv.API().Groups.Delete(context.Background(), g.Id) })
+		return g
+	}
+	gUX := mkGroup("e2e-mg-union-x")
+	gUY := mkGroup("e2e-mg-union-y")
+	gMP := mkGroup("e2e-mg-mix-p")
+	gMQ := mkGroup("e2e-mg-mix-q")
+
+	ephemeral := false
+	sk, err := srv.API().SetupKeys.Create(ctx, api.PostApiSetupKeysJSONRequestBody{
+		Name:       "e2e-mg-client",
+		Type:       "reusable",
+		ExpiresIn:  86400,
+		UsageLimit: 0,
+		AutoGroups: []string{gUX.Id, gUY.Id, gMP.Id, gMQ.Id}, // client in all four groups
+		Ephemeral:  &ephemeral,
+	})
+	require.NoError(t, err, "mint setup key")
+	require.NotEmpty(t, sk.Key, "setup key plaintext")
+
+	staticKey := "static-e2e-token"
+	enabled := true
+
+	// P1 — union scenario: two restricting policies, one per group.
+	p1, err := srv.CreateProvider(ctx, api.AgentNetworkProviderRequest{
+		Name:             "e2e-mg-union",
+		ProviderId:       "openai_api",
+		UpstreamUrl:      vllm.URL,
+		ApiKey:           &staticKey,
+		Enabled:          ptr(true),
+		Models:           priced(unionA, unionB, unionC),
+		BootstrapCluster: ptr(harness.AgentNetworkCluster),
+	})
+	require.NoError(t, err, "create union provider")
+	t.Cleanup(func() { _ = srv.DeleteProvider(context.Background(), p1.Id) })
+
+	polUX, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-mg-union-x",
+		Enabled:                &enabled,
+		SourceGroups:           []string{gUX.Id},
+		DestinationProviderIds: []string{p1.Id},
+		GuardrailIds:           &[]string{mkAllowGuardrail(t, ctx, "e2e-mg-union-x", unionA).Id},
+	})
+	require.NoError(t, err, "create union policy X")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polUX.Id) })
+
+	polUY, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-mg-union-y",
+		Enabled:                &enabled,
+		SourceGroups:           []string{gUY.Id},
+		DestinationProviderIds: []string{p1.Id},
+		GuardrailIds:           &[]string{mkAllowGuardrail(t, ctx, "e2e-mg-union-y", unionB).Id},
+	})
+	require.NoError(t, err, "create union policy Y")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polUY.Id) })
+
+	// P2 — mixed scenario: one restricting policy + one un-guardrailed policy.
+	p2, err := srv.CreateProvider(ctx, api.AgentNetworkProviderRequest{
+		Name:        "e2e-mg-mix",
+		ProviderId:  "openai_api",
+		UpstreamUrl: vllm.URL,
+		ApiKey:      &staticKey,
+		Enabled:     ptr(true),
+		Models:      priced(mixA, mixZ),
+	})
+	require.NoError(t, err, "create mix provider")
+	t.Cleanup(func() { _ = srv.DeleteProvider(context.Background(), p2.Id) })
+
+	polMP, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-mg-mix-p",
+		Enabled:                &enabled,
+		SourceGroups:           []string{gMP.Id},
+		DestinationProviderIds: []string{p2.Id},
+		GuardrailIds:           &[]string{mkAllowGuardrail(t, ctx, "e2e-mg-mix-p", mixA).Id},
+	})
+	require.NoError(t, err, "create mix policy P")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polMP.Id) })
+
+	polMQ, err := srv.CreatePolicy(ctx, api.AgentNetworkPolicyRequest{
+		Name:                   "e2e-mg-mix-q",
+		Enabled:                &enabled,
+		SourceGroups:           []string{gMQ.Id},
+		DestinationProviderIds: []string{p2.Id}, // NO guardrail -> unrestricted
+	})
+	require.NoError(t, err, "create mix policy Q")
+	t.Cleanup(func() { _ = srv.DeletePolicy(context.Background(), polMQ.Id) })
+
+	settings, err := srv.GetSettings(ctx)
+	require.NoError(t, err, "read settings")
+	require.NotEmpty(t, settings.Endpoint, "endpoint must be assigned")
+
+	proxyToken, err := srv.CreateProxyTokenCLI(ctx, "e2e-mg-proxy")
+	require.NoError(t, err, "mint proxy token")
+	px, err := harness.StartProxy(ctx, srv, proxyToken)
+	require.NoError(t, err, "start proxy")
+	t.Cleanup(func() { _ = px.Terminate(context.Background()) })
+
+	cl, err := harness.StartClient(ctx, srv, sk.Key)
+	require.NoError(t, err, "start client")
+	t.Cleanup(func() { _ = cl.Terminate(context.Background()) })
+
+	require.NoError(t, cl.WaitConnected(ctx, 90*time.Second), "client must connect to management")
+	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
+	require.NoError(t, err, "resolve endpoint to proxy IP")
+	if err := cl.WaitProxyPeer(ctx, 180*time.Second); err != nil {
+		t.Fatalf("client did not see the proxy peer: %v\n=== proxy logs ===\n%s", err, px.Logs(context.Background()))
+	}
+
+	send := func(model string) (int, string) {
+		code, body, cerr := cl.Chat(ctx, settings.Endpoint, proxyIP, harness.WireChat, model, "Reply with exactly: pong", "")
+		require.NoError(t, cerr, "request must reach the proxy")
+		return code, body
+	}
+	sendUntil200 := func(model string) (int, string) {
+		var code int
+		var body string
+		deadline := time.Now().Add(90 * time.Second)
+		for time.Now().Before(deadline) {
+			code, body = send(model)
+			if code == 200 {
+				break
+			}
+			time.Sleep(5 * time.Second)
+		}
+		return code, body
+	}
+
+	t.Run("union across the user's groups", func(t *testing.T) {
+		code, body := sendUntil200(unionA)
+		assert.Equal(t, 200, code, "model allowed by group X must be served; body: %s\n=== proxy logs ===\n%s", body, px.Logs(context.Background()))
+
+		code, body = sendUntil200(unionB)
+		assert.Equal(t, 200, code, "model allowed by group Y must also be served (union across the user's groups); body: %s", body)
+
+		code, body = send(unionC)
+		assert.Equal(t, 403, code, "a model on neither group's allowlist must be denied; body: %s", body)
+		assert.Contains(t, body, "llm_policy.model_blocked")
+	})
+
+	t.Run("an un-guardrailed group lifts the restriction", func(t *testing.T) {
+		code, body := sendUntil200(mixA)
+		assert.Equal(t, 200, code, "the restricted group's model must be served; body: %s\n=== proxy logs ===\n%s", body, px.Logs(context.Background()))
+
+		code, body = sendUntil200(mixZ)
+		assert.Equal(t, 200, code,
+			"a non-allowlisted model must be served because the user is also in a group whose policy has no guardrail; body: %s\n=== proxy logs ===\n%s", body, px.Logs(context.Background()))
+	})
+}
+
+// mkAllowGuardrail creates a guardrail whose model allowlist is enabled and holds
+// exactly the given model, registering cleanup.
+func mkAllowGuardrail(t *testing.T, ctx context.Context, name, model string) api.AgentNetworkGuardrail {
+	t.Helper()
+	var gr api.AgentNetworkGuardrailRequest
+	gr.Name = name
+	gr.Checks.ModelAllowlist.Enabled = true
+	gr.Checks.ModelAllowlist.Models = []string{model}
+	g, err := srv.CreateGuardrail(ctx, gr)
+	require.NoError(t, err, "create guardrail %s", name)
+	t.Cleanup(func() { _ = srv.DeleteGuardrail(context.Background(), g.Id) })
+	return g
+}

e2e/agentnetwork/guardrail_test.go

@@ -4,6 +4,7 @@ package agentnetwork
 
 import (
 	"context"
+	"regexp"
 	"strings"
 	"testing"
 	"time"
@@ -15,13 +16,29 @@ import (
 	"github.com/netbirdio/netbird/shared/management/http/api"
 )
 
+// bedrockRegionPrefixes and bedrockVersionSuffix mirror the proxy's Bedrock
+// model normalization (region/inference-profile prefix + version suffix) so the
+// provider is registered under the same catalog key the router matches against.
+var (
+	bedrockRegionPrefixes = []string{"us.", "eu.", "apac.", "global."}
+	bedrockVersionSuffix  = regexp.MustCompile(`-(\d{8}-)?v\d+(:\d+)?$`)
+)
+
 // catalogModel returns the normalized catalog id the proxy stamps for a
-// path-routed provider's configured model — the form the guardrail allowlist is
-// compared against (region prefix / @version stripped).
+// path-routed provider's configured model — the form the router and guardrail
+// allowlist compare against (Bedrock region prefix + version stripped, Vertex
+// @version stripped).
 func catalogModel(pc providerCase) string {
 	switch pc.kind {
 	case harness.WireBedrock:
-		return strings.TrimPrefix(pc.model, "us.")
+		m := pc.model
+		for _, p := range bedrockRegionPrefixes {
+			if strings.HasPrefix(m, p) {
+				m = m[len(p):]
+				break
+			}
+		}
+		return bedrockVersionSuffix.ReplaceAllString(m, "")
 	case harness.WireVertex:
 		return strings.SplitN(pc.model, "@", 2)[0]
 	default:
@@ -35,7 +52,9 @@ func catalogModel(pc providerCase) string {
 func disallowedModel(pc providerCase) string {
 	switch pc.kind {
 	case harness.WireBedrock:
-		return "us.anthropic.claude-opus-4-8"
+		// Same profile prefix as the allowed model so only the model name
+		// differs; the guardrail must deny it before it reaches AWS.
+		return strings.SplitN(pc.model, ".", 2)[0] + ".anthropic.claude-opus-4-8"
 	case harness.WireVertex:
 		return "claude-opus-4-8@20250101"
 	default:
@@ -55,7 +74,7 @@ func sendModel(ctx context.Context, t *testing.T, cl *harness.Client, endpoint,
 	case harness.WireVertex:
 		code, _, err = cl.Vertex(ctx, endpoint, proxyIP, pc.project, pc.region, model, "Reply with exactly: pong", "")
 	default:
-		code, _, err = cl.Chat(ctx, endpoint, proxyIP, pc.kind, model, "Reply with exactly: pong", "")
+		code, _, err = cl.ChatPrefixed(ctx, endpoint, proxyIP, pc.pathPrefix, pc.kind, model, "Reply with exactly: pong", "")
 	}
 	require.NoError(t, err, "request must reach the proxy for %s", pc.name)
 	return code
@@ -147,11 +166,12 @@ func TestModelAllowlistEnforced(t *testing.T) {
 	t.Cleanup(func() { _ = cl.Terminate(context.Background()) })
 
 	require.NoError(t, cl.WaitConnected(ctx, 90*time.Second), "client must connect to management")
+	// Probe first: the GET resolves the endpoint (DNS error fails) and its first packet wakes the lazy proxy peer, so WaitProxyPeer sees it connected; any HTTP status counts.
+	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
+	require.NoError(t, err, "resolve agent-network endpoint to proxy IP")
 	if err := cl.WaitProxyPeer(ctx, 180*time.Second); err != nil {
 		t.Fatalf("client did not see the proxy peer: %v\n=== proxy logs ===\n%s", err, px.Logs(context.Background()))
 	}
-	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
-	require.NoError(t, err, "resolve agent-network endpoint to proxy IP")
 
 	for _, pc := range providers {
 		pc := pc

e2e/agentnetwork/skiptls_test.go

@@ -104,11 +104,12 @@ func TestProviderSkipTLSVerification(t *testing.T) {
 	t.Cleanup(func() { _ = cl.Terminate(context.Background()) })
 
 	require.NoError(t, cl.WaitConnected(ctx, 90*time.Second), "client must connect to management")
+	// Probe first: the GET resolves the endpoint (DNS error fails) and its first packet wakes the lazy proxy peer, so WaitProxyPeer sees it connected; any HTTP status counts.
+	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
+	require.NoError(t, err, "resolve endpoint to proxy IP")
 	if err := cl.WaitProxyPeer(ctx, 180*time.Second); err != nil {
 		t.Fatalf("client did not see the proxy peer: %v\n=== proxy logs ===\n%s", err, px.Logs(context.Background()))
 	}
-	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
-	require.NoError(t, err, "resolve endpoint to proxy IP")
 
 	// Positive: skip=true reaches the self-signed upstream. Retry to absorb
 	// tunnel/DNS jitter on the first call; success also proves the path works.

e2e/agentnetwork/vllm_test.go

@@ -106,11 +106,12 @@ func TestVLLMProvider(t *testing.T) {
 	t.Cleanup(func() { _ = cl.Terminate(context.Background()) })
 
 	require.NoError(t, cl.WaitConnected(ctx, 90*time.Second), "client must connect to management")
+	// Probe first: the GET resolves the endpoint (DNS error fails) and its first packet wakes the lazy proxy peer, so WaitProxyPeer sees it connected; any HTTP status counts.
+	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
+	require.NoError(t, err, "resolve endpoint to proxy IP")
 	if err := cl.WaitProxyPeer(ctx, 180*time.Second); err != nil {
 		t.Fatalf("client did not see the proxy peer: %v\n=== proxy logs ===\n%s", err, px.Logs(context.Background()))
 	}
-	proxyIP, err := cl.ResolveProxyIP(ctx, settings.Endpoint)
-	require.NoError(t, err, "resolve endpoint to proxy IP")
 
 	before, _ := srv.ListAccessLogs(ctx)
 	sessionID := "e2e-session-vllm"

e2e/harness/client.go

@@ -4,6 +4,7 @@ package harness
 
 import (
 	"context"
+	"errors"
 	"fmt"
 	"io"
 	"os/exec"
@@ -167,22 +168,54 @@ func (cl *Client) pollStatus(ctx context.Context, timeout time.Duration, want st
 	return fmt.Errorf("timed out waiting for %q; last status:\n%s", want, last)
 }
 
-// ResolveProxyIP resolves the agent-network endpoint to the proxy peer's
-// NetBird IP from inside the client (via magic DNS).
+const (
+	// curlExitCouldNotResolve is curl's exit code for a DNS resolution failure, distinct from connection-level failures.
+	curlExitCouldNotResolve = 6
+	// dnsProbeRetryWindow bounds DNS-failure retries: the synthesized zone lands a beat after management connects, so early NXDOMAIN is propagation; a zone still absent after this window is a real failure.
+	dnsProbeRetryWindow   = 30 * time.Second
+	dnsProbeRetryInterval = 2 * time.Second
+)
+
+// ResolveProxyIP GETs https://<endpoint>/ from the client's netns: any HTTP status proves DNS + tunnel and wakes the lazy proxy peer; only DNS failures retry, within dnsProbeRetryWindow. Returns the connected IP for --resolve pinning.
 func (cl *Client) ResolveProxyIP(ctx context.Context, endpoint string) (string, error) {
-	code, reader, err := cl.container.Exec(ctx, []string{"getent", "hosts", endpoint}, tcexec.Multiplexed())
-	if err != nil {
-		return "", err
+	args := []string{
+		"run", "--rm",
+		"--network", "container:" + cl.container.GetContainerID(),
+		curlImage,
+		"-ksS", "-o", "/dev/null",
+		"--connect-timeout", "30", "--max-time", "60",
+		"-w", "%{remote_ip}",
+		"https://" + endpoint + "/",
 	}
-	out, _ := io.ReadAll(reader)
-	if code != 0 {
-		return "", fmt.Errorf("getent hosts %s exited %d", endpoint, code)
+	deadline := time.Now().Add(dnsProbeRetryWindow)
+	for {
+		cmd := exec.CommandContext(ctx, "docker", args...)
+		var stdout, stderr strings.Builder
+		cmd.Stdout = &stdout
+		cmd.Stderr = &stderr
+		err := cmd.Run()
+		if err == nil {
+			ip := strings.TrimSpace(stdout.String())
+			if ip == "" {
+				return "", fmt.Errorf("got an HTTP response from %s but no remote IP", endpoint)
+			}
+			return ip, nil
+		}
+
+		var exitErr *exec.ExitError
+		if !errors.As(err, &exitErr) || exitErr.ExitCode() != curlExitCouldNotResolve {
+			return "", fmt.Errorf("no HTTP response from %s: %w (%s)", endpoint, err, strings.TrimSpace(stderr.String()))
+		}
+		dnsErr := fmt.Errorf("DNS resolution failed for %s: %s", endpoint, strings.TrimSpace(stderr.String()))
+		if time.Until(deadline) < dnsProbeRetryInterval {
+			return "", dnsErr
+		}
+		select {
+		case <-ctx.Done():
+			return "", fmt.Errorf("%w (%w)", dnsErr, ctx.Err())
+		case <-time.After(dnsProbeRetryInterval):
+		}
 	}
-	fields := strings.Fields(string(out))
-	if len(fields) == 0 {
-		return "", fmt.Errorf("no address for %s", endpoint)
-	}
-	return fields[0], nil
 }
 
 // Wire shapes for Chat.
@@ -206,18 +239,29 @@ const (
 // the wire shape: WireChat (OpenAI) or WireMessages (Anthropic). A non-empty
 // sessionID is sent as the universal x-session-id header the proxy records.
 func (cl *Client) Chat(ctx context.Context, endpoint, proxyIP, kind, model, prompt, sessionID string) (int, string, error) {
+	return cl.ChatPrefixed(ctx, endpoint, proxyIP, "", kind, model, prompt, sessionID)
+}
+
+// ChatPrefixed is Chat with a base-URL path prefix prepended to the wire
+// path, mirroring agents whose base URL carries a shape-selecting prefix that
+// rides through to the upstream — e.g. Claude Code against a Kimi provider
+// sets ANTHROPIC_BASE_URL=https://<endpoint>/anthropic so the proxy forwards
+// /anthropic/v1/messages to Moonshot's Anthropic surface while the provider's
+// upstream URL stays the bare https://api.moonshot.ai. Empty prefix is plain
+// Chat.
+func (cl *Client) ChatPrefixed(ctx context.Context, endpoint, proxyIP, pathPrefix, kind, model, prompt, sessionID string) (int, string, error) {
 	var path, body string
 	var headers []string
 	switch kind {
 	case WireMessages:
 		path = "/v1/messages"
 		headers = []string{"anthropic-version: 2023-06-01"}
-		body = fmt.Sprintf(`{"model":%q,"max_tokens":64,"messages":[{"role":"user","content":%q}]}`, model, prompt)
+		body = fmt.Sprintf(`{"model":%q,"max_tokens":2048,"messages":[{"role":"user","content":%q}]}`, model, prompt)
 	default:
 		path = "/v1/chat/completions"
 		body = fmt.Sprintf(`{"model":%q,"messages":[{"role":"user","content":%q}]}`, model, prompt)
 	}
-	return cl.post(ctx, endpoint, proxyIP, path, body, withSessionID(headers, sessionID))
+	return cl.post(ctx, endpoint, proxyIP, pathPrefix+path, body, withSessionID(headers, sessionID))
 }
 
 // Vertex issues an Anthropic-on-Vertex rawPredict POST over the tunnel. Unlike
@@ -227,7 +271,7 @@ func (cl *Client) Chat(ctx context.Context, endpoint, proxyIP, kind, model, prom
 // is sent as the universal x-session-id header the proxy records.
 func (cl *Client) Vertex(ctx context.Context, endpoint, proxyIP, project, region, model, prompt, sessionID string) (int, string, error) {
 	path := fmt.Sprintf("/v1/projects/%s/locations/%s/publishers/anthropic/models/%s:rawPredict", project, region, model)
-	body := fmt.Sprintf(`{"anthropic_version":"vertex-2023-10-16","max_tokens":64,"messages":[{"role":"user","content":%q}]}`, prompt)
+	body := fmt.Sprintf(`{"anthropic_version":"vertex-2023-10-16","max_tokens":2048,"messages":[{"role":"user","content":%q}]}`, prompt)
 	return cl.post(ctx, endpoint, proxyIP, path, body, withSessionID(nil, sessionID))
 }
 
@@ -238,7 +282,7 @@ func (cl *Client) Vertex(ctx context.Context, endpoint, proxyIP, project, region
 // header the proxy records.
 func (cl *Client) Bedrock(ctx context.Context, endpoint, proxyIP, model, prompt, sessionID string) (int, string, error) {
 	path := "/model/" + model + "/invoke"
-	body := fmt.Sprintf(`{"anthropic_version":"bedrock-2023-05-31","max_tokens":64,"messages":[{"role":"user","content":%q}]}`, prompt)
+	body := fmt.Sprintf(`{"anthropic_version":"bedrock-2023-05-31","max_tokens":2048,"messages":[{"role":"user","content":%q}]}`, prompt)
 	return cl.post(ctx, endpoint, proxyIP, path, body, withSessionID(nil, sessionID))
 }
 
@@ -297,7 +341,7 @@ func (cl *Client) Terminate(ctx context.Context) error {
 	return cl.container.Terminate(ctx)
 }
 
-// containerLogs reads up to 256 KiB of a container's logs for diagnostics.
+// containerLogs reads up to 4 MiB of a container's logs for diagnostics — enough for a whole provider-matrix run.
 func containerLogs(ctx context.Context, c testcontainers.Container) string {
 	if c == nil {
 		return ""
@@ -307,6 +351,6 @@ func containerLogs(ctx context.Context, c testcontainers.Container) string {
 		return fmt.Sprintf("<logs error: %v>", err)
 	}
 	defer r.Close()
-	b, _ := io.ReadAll(io.LimitReader(r, 256<<10))
+	b, _ := io.ReadAll(io.LimitReader(r, 4<<20))
 	return string(b)
 }

e2e/harness/combined.go

@@ -221,6 +221,29 @@ func (c *Combined) CreateProxyTokenCLI(ctx context.Context, name string) (string
 	return "", fmt.Errorf("token not found in CLI output: %s", string(out))
 }
 
+// SnapshotStoreDB copies the management sqlite store (with WAL/SHM sidecars) out of the bind-mounted
+// data dir into dstDir and returns the copy's path; reading a copy avoids locking against live writes.
+func (c *Combined) SnapshotStoreDB(dstDir string) (string, error) {
+	src := filepath.Join(c.workDir, "data", "store.db")
+	if _, err := os.Stat(src); err != nil {
+		return "", fmt.Errorf("management store not found at %s: %w", src, err)
+	}
+	dst := filepath.Join(dstDir, "store.db")
+	for _, suffix := range []string{"", "-wal", "-shm"} {
+		data, err := os.ReadFile(src + suffix)
+		if err != nil {
+			if os.IsNotExist(err) && suffix != "" {
+				continue // sidecar only exists in WAL mode
+			}
+			return "", fmt.Errorf("read %s: %w", src+suffix, err)
+		}
+		if err := os.WriteFile(dst+suffix, data, 0o600); err != nil {
+			return "", fmt.Errorf("write %s: %w", dst+suffix, err)
+		}
+	}
+	return dst, nil
+}
+
 // Logs returns the combined server container logs, for diagnostics.
 func (c *Combined) Logs(ctx context.Context) string {
 	return containerLogs(ctx, c.container)

e2e/harness/proxy.go

@@ -38,7 +38,11 @@ type Proxy struct {
 // network, registered via the given account proxy token and serving the
 // AgentNetworkCluster over a self-signed wildcard cert. It does not wait for
 // peer connectivity — callers poll management for the proxy peer.
-func StartProxy(ctx context.Context, c *Combined, proxyToken string) (*Proxy, error) {
+// StartProxy launches the reverse-proxy container. Optional envOverrides are
+// merged into the container environment after the defaults, so callers can set
+// or override any NB_PROXY_* var (e.g. NB_PROXY_TUNNEL_CACHE_TTL for tests that
+// need a short authorization-cache window).
+func StartProxy(ctx context.Context, c *Combined, proxyToken string, envOverrides ...map[string]string) (*Proxy, error) {
 	root, err := repoRoot()
 	if err != nil {
 		return nil, err
@@ -93,6 +97,12 @@ func StartProxy(ctx context.Context, c *Combined, proxyToken string) (*Proxy, er
 		WaitingFor: wait.ForLog("Initial mapping sync complete").WithStartupTimeout(90 * time.Second),
 	}
 
+	for _, ov := range envOverrides {
+		for k, v := range ov {
+			req.Env[k] = v
+		}
+	}
+
 	ctr, err := testcontainers.GenericContainer(ctx, testcontainers.GenericContainerRequest{
 		ContainerRequest: req,
 		Started:          true,

idp/dex/config.go

@@ -308,7 +308,7 @@ func (s *Storage) OpenStorage(logger *slog.Logger) (storage.Storage, error) {
 		if file == "" {
 			return nil, fmt.Errorf("sqlite3 storage requires 'file' config")
 		}
-		return (&sql.SQLite3{File: file}).Open(logger)
+		return newSQLite3(file).Open(logger)
 	case "postgres":
 		dsn, _ := s.Config["dsn"].(string)
 		if dsn == "" {