fix(inbound): refresh socket.io auth on (re)connect via callback

Backend issues short-lived guildAccessToken (TTL=900s). The previous
`auth: { token: tok }` shape captured the JWT once in connectAgent's
closure: after socket.io's auto-reconnect the backend kept getting the
same expired JWT and silently rejected the handshake at the application
layer (RealtimeGateway logs 'socket rejected: <id>'). The client's
'connect' event still fired (TCP succeeded) so the plugin happily ran
the channel-resync, emitted join_channel into the void, and logged
'joined N channel(s)' while the backend was actually broadcasting
message.created to a room with zero subscribers. End-user symptom:
DMs/group messages to agents silently dropped 15 min after gateway
start, with no error anywhere on the agent side.

Switch to the callback form, which socket.io re-evaluates on every
(re)connect — same call site we already use for the HTTP path via
freshGuildToken/tokenCache.

Verified in sim (commit 2acb084 + this patch):
1. Connect new DM channel + post msg -> dispatch + reply ✓
2. `docker restart fabric-backend-guild` to force socket disconnect
3. Plugin reconnects automatically and logs
   'fabric: agent recruiter joined 12 channel(s) on sim-guild-1' ✓
   (without the fix this reconnect was silently rejected; sim used to
    log 'WARN socket rejected: <id>' on the guild backend)

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>

dist/fabric/index.js

@@ -94,7 +94,7 @@ export default defineChannelPluginEntry({
             void inbound.start().then(() => {
                 if (!inbound)
                     return;
-                presence = new PresenceSync(api.logger);
+                presence = new PresenceSync(api.logger, client);
                 presence.setAccounts(inbound.getPresenceAccounts());
                 presence.start();
                 api.logger.info(`fabric: presence-sync started for ${inbound.getPresenceAccounts().length} account(s)`);

dist/fabric/src/channel.js

@@ -11,6 +11,15 @@
 import { createChatChannelPlugin, createChannelPluginBase, buildChannelOutboundSessionRoute, } from 'openclaw/plugin-sdk/core';
 import { FabricClient } from './fabric-client.js';
 import { listFabricAccountIds, resolveFabricAccount, resolveDefaultFabricAccountId, } from './accounts.js';
+import { getChannelType } from './channel-meta.js';
+export function fabricPeerRoutingForXType(xType) {
+    if (xType === 'dm')
+        return { peerKind: 'direct', chatType: 'direct' };
+    return { peerKind: 'group', chatType: 'group' };
+}
+export function fabricPeerRoutingForChannel(channelId) {
+    return fabricPeerRoutingForXType(getChannelType(channelId));
+}
 // ---- target grammar: fabric:<channelId> ----
 export function stripFabricTargetPrefix(raw) {
     let s = (raw ?? '').trim();
@@ -38,13 +47,18 @@ export function resolveFabricOutboundSessionRoute(params) {
     const id = stripFabricTargetPrefix(params.target);
     if (!id)
         return null;
+    // Consult the channel-meta cache populated by inbound — DM channels
+    // need peer.kind='direct' so the outbound session key matches the
+    // inbound one. Cache miss falls back to 'group' (the pre-fix default,
+    // no regression on cold cache).
+    const { peerKind, chatType } = fabricPeerRoutingForChannel(id);
     return buildChannelOutboundSessionRoute({
         cfg: params.cfg,
         agentId: params.agentId,
         channel: 'fabric',
         accountId: params.accountId,
-        peer: { kind: 'group', id },
+        peer: { kind: peerKind, id },
-        chatType: 'group',
+        chatType,
         from: `fabric:channel:${id}`,
         to: `fabric:${id}`,
     });

dist/fabric/src/inbound.js

@@ -4,6 +4,7 @@ import { join } from 'node:path';
 import { io } from 'socket.io-client';
 import { dispatchInboundReplyWithBase } from 'openclaw/plugin-sdk/inbound-reply-dispatch';
 import { resolveCoalesce } from './accounts.js';
+import { fabricPeerRoutingForXType } from './channel.js';
 import { recordChannelType } from './channel-meta.js';
 import { enqueueDelivery, flushFabricForChannel } from './coalesce.js';
 export class FabricInbound {
@@ -227,38 +228,59 @@ export class FabricInbound {
         for (const entry of this.identity.list()) {
             if (!entry.fabricUserId)
                 continue;
-            const presenceGuildUrl = this.firstGuildEndpointByAgent.get(entry.agentId);
+            const presenceGuild = this.firstGuildByAgent.get(entry.agentId);
-            if (!presenceGuildUrl)
+            if (!presenceGuild)
                 continue;
             out.push({
                 agentId: entry.agentId,
                 fabricUserId: entry.fabricUserId,
-                guildBaseUrl: presenceGuildUrl,
+                guildBaseUrl: presenceGuild.endpoint,
+                guildNodeId: presenceGuild.nodeId,
                 fabricApiKey: entry.fabricApiKey,
             });
         }
         return out;
     }
     // Filled by connectAgent for each (agent, guild). Tracks ONLY the first
-    // guild per agent (used as the presence-push target).
+    // guild per agent (used as the presence-push target). Stores both
-    firstGuildEndpointByAgent = new Map();
+    // endpoint and nodeId — presence-sync needs both: endpoint to build
+    // the URL, nodeId to pick the matching guildAccessToken from a fresh
+    // agent-login response.
+    firstGuildByAgent = new Map();
     async connectAgent(agentId, session) {
         const selfUserId = session.user.id;
         // First-guild capture for presence-sync push target. session.guilds is
         // already in priority order from Center; we take the first one with a
         // valid endpoint and stop. Multi-guild presence is a future concern.
-        if (!this.firstGuildEndpointByAgent.has(agentId)) {
+        if (!this.firstGuildByAgent.has(agentId)) {
             const firstGuild = session.guilds.find((g) => typeof g.endpoint === 'string' && g.endpoint.length > 0);
             if (firstGuild)
-                this.firstGuildEndpointByAgent.set(agentId, firstGuild.endpoint);
+                this.firstGuildByAgent.set(agentId, { endpoint: firstGuild.endpoint, nodeId: firstGuild.nodeId });
         }
         for (const g of session.guilds) {
             const tok = session.guildAccessTokens.find((t) => t.guildNodeId === g.nodeId)?.token;
             if (!tok)
                 continue;
+            // Use the *callback* form of `auth` so socket.io re-evaluates the JWT
+            // on every (re)connect. The single-shot `auth: { token: tok }` shape
+            // captured the token in closure: after socket.io's silent auto-reconnect
+            // the backend got the same JWT that expired ~15 min into the session
+            // (guildAccessToken TTL = 900s) and silently rejected the handshake at
+            // the application layer. The client's `connect` event still fired (TCP
+            // succeeded), so the plugin happily ran the channel-resync, emitted
+            // `join_channel` into the void, and logged "joined N channel(s)" while
+            // the backend was actually broadcasting message.created to a room with
+            // zero subscribers. End user symptom: DMs to agents silently dropped.
             const socket = io(`${g.endpoint}/realtime`, {
                 transports: ['websocket'],
-                auth: { token: tok },
+                auth: (cb) => {
+                    // Best-effort fresh token; on transient failure fall back to the
+                    // last known good one. tokenCache also keeps HTTP calls (attachment
+                    // download / reply post) from 401'ing in the same window.
+                    this.freshGuildToken(agentId, g.nodeId, session)
+                        .then((fresh) => cb({ token: fresh ?? tok }))
+                        .catch(() => cb({ token: tok }));
+                },
                 autoConnect: false,
             });
             // Tracked socket.io rooms for this (agent, guild). The initial fetch
@@ -433,11 +455,19 @@ export class FabricInbound {
         const core = this.core;
         const cfg = this.cfg;
         try {
+            // Route by xType. DM channels need peer.kind='direct' so openclaw
+            // treats them as 1:1 (sessionKey 'agent:<id>:fabric:direct:<chan>'
+            // and ctx.ChatType='direct') rather than as a multi-party group.
+            // Without this, the agent's user-prompt metadata says
+            // 'is_group_chat: true' on a DM and downstream prompt logic
+            // (commands-handlers `isDirectMessage` checks ChatType==='direct')
+            // misclassifies the turn.
+            const { peerKind, chatType } = fabricPeerRoutingForXType(m.xType);
             const route = core.channel.routing.resolveAgentRoute({
                 cfg: this.cfg,
                 channel: 'fabric',
                 accountId: agentId,
-                peer: { kind: 'group', id: channelId },
+                peer: { kind: peerKind, id: channelId },
             });
             const storePath = core.channel.session.resolveStorePath(cfg.session?.store, {
                 agentId: route.agentId,
@@ -451,7 +481,7 @@ export class FabricInbound {
                 To: `fabric:${channelId}`,
                 SessionKey: route.sessionKey,
                 AccountId: route.accountId ?? agentId,
-                ChatType: 'group',
+                ChatType: chatType,
                 ConversationLabel: `fabric:${guild.nodeId}`,
                 SenderId: m.authorUserId ?? 'fabric',
                 Provider: 'fabric',

dist/fabric/src/presence-sync.js

@@ -1,26 +1,25 @@
-/**
+// Guild access JWTs expire every 900s. Refresh ~2 min early to stay
- * presence-sync — read each connected agent's HF status (via the
+// safely inside the window even if a tick runs late.
- * cross-plugin `globalThis.__hfAgentStatus.get(agentId)` exposed by
+const TOKEN_TTL_MS = (15 - 2) * 60 * 1000;
- * HarborForge.OpenclawPlugin) and push diffs to Fabric.Backend.Guild
- * `PUT /agents/:userId/presence` so the backend can apply busy-discard
- * on `announce`-type channel deliveries.
- *
- * Push model: we only PUT when an agent's status actually changes
- * (since the last push). The HF-side accessor has its own TTL cache
- * to absorb the every-30s polling.
- *
- * If HF plugin isn't loaded (`__hfAgentStatus` undefined), the loop
- * is a no-op — Fabric backend defaults presence to 'unknown' which is
- * treated as not-busy. Announce-channel delivery still works; busy
- * filtering simply doesn't kick in.
- */
 export class PresenceSync {
     logger;
+    client;
     timer = null;
     lastStatus = new Map(); // by agentId
     accounts = new Map();
-    constructor(logger) {
+    tokenCache = new Map(); // by agentId
+    // Mutex flag: a tick iterates accounts serially with `await` on each
+    // agent-login + PUT round-trip, so a single tick can easily run 20+s
+    // when there are many accounts. setInterval(intervalMs) does NOT wait
+    // for the previous tick to finish — without this guard the next tick
+    // fires on top of a still-running one and two parallel iterations
+    // PUT the same agentId within milliseconds. That tipped the backend's
+    // first-time-insert race (separate fix in Fabric.Backend.Guild) into
+    // 500s on prod. Guarded ticks just skip a beat instead.
+    inflight = false;
+    constructor(logger, client) {
         this.logger = logger;
+        this.client = client;
     }
     setAccounts(accounts) {
         this.accounts.clear();
@@ -42,7 +41,49 @@ export class PresenceSync {
             this.timer = null;
         }
     }
+    /**
+     * Fetch a fresh guildAccessToken for `acct`, caching it under the
+     * agentId until just before its JWT expiry. Returns null on login
+     * failure or if the session has no matching guild — caller logs +
+     * skips the PUT.
+     */
+    async ensureGuildToken(acct) {
+        const now = Date.now();
+        const cached = this.tokenCache.get(acct.agentId);
+        if (cached && cached.expiresAt > now)
+            return cached.token;
+        let session;
+        try {
+            session = await this.client.agentLogin(acct.fabricApiKey);
+        }
+        catch (err) {
+            this.logger.warn(`fabric: presence-sync agent-login failed for ${acct.agentId}: ${String(err)}`);
+            return null;
+        }
+        const entry = session.guildAccessTokens.find((g) => g.guildNodeId === acct.guildNodeId);
+        if (!entry?.token) {
+            this.logger.warn(`fabric: presence-sync no guild token for ${acct.agentId} guild=${acct.guildNodeId}`);
+            return null;
+        }
+        this.tokenCache.set(acct.agentId, { token: entry.token, expiresAt: now + TOKEN_TTL_MS });
+        return entry.token;
+    }
     async tick() {
+        // Mutex: see the `inflight` field declaration for the why. Drop
+        // overlapping ticks rather than letting them run concurrently —
+        // status is gated by `lastStatus !== bridge.get`, so skipping a
+        // beat costs nothing the next beat won't catch.
+        if (this.inflight)
+            return;
+        this.inflight = true;
+        try {
+            await this.tickInner();
+        }
+        finally {
+            this.inflight = false;
+        }
+    }
+    async tickInner() {
         const bridge = globalThis['__hfAgentStatus'];
         if (!bridge || typeof bridge.get !== 'function')
             return; // HF plugin not loaded — skip
@@ -58,13 +99,22 @@ export class PresenceSync {
                 continue;
             if (this.lastStatus.get(agentId) === status)
                 continue; // no change → no PUT
+            const guildToken = await this.ensureGuildToken(acct);
+            if (!guildToken)
+                continue;
             try {
-                const url = `${acct.guildBaseUrl.replace(/\/$/, '')}/agents/${encodeURIComponent(acct.fabricUserId)}/presence`;
+                // Endpoint: PUT /api/agents/:userId/presence. ApiKeyGuard (global
+                // APP_GUARD) requires `Authorization: Bearer <guildAccessToken>`
+                // — NOT the agent's raw fabricApiKey. Pre-v1: this loop sent
+                // x-api-key and got 401 "missing bearer token" forever. The /api
+                // prefix is required because the guild backend sets a global
+                // 'api' prefix in main.ts setGlobalPrefix('api').
+                const url = `${acct.guildBaseUrl.replace(/\/$/, '')}/api/agents/${encodeURIComponent(acct.fabricUserId)}/presence`;
                 const res = await fetch(url, {
                     method: 'PUT',
                     headers: {
                         'content-type': 'application/json',
-                        'x-api-key': acct.fabricApiKey,
+                        authorization: `Bearer ${guildToken}`,
                     },
                     body: JSON.stringify({ status, source: 'hf-plugin' }),
                 });
@@ -73,6 +123,11 @@ export class PresenceSync {
                     this.logger.info(`fabric: presence-sync ${agentId} → ${status}`);
                 }
                 else {
+                    // 401 here usually means the cached token went stale unexpectedly
+                    // (server-side rotation or clock skew) — drop the cache so the
+                    // next tick re-logs-in.
+                    if (res.status === 401)
+                        this.tokenCache.delete(agentId);
                     this.logger.warn(`fabric: presence-sync PUT ${agentId} failed: ${res.status}`);
                 }
             }

index.ts

@@ -116,7 +116,7 @@ export default defineChannelPluginEntry({
       // their fabricUserId + first guild endpoint populated).
       void inbound.start().then(() => {
         if (!inbound) return;
-        presence = new PresenceSync(api.logger);
+        presence = new PresenceSync(api.logger, client);
         presence.setAccounts(inbound.getPresenceAccounts());
         presence.start();
         api.logger.info(`fabric: presence-sync started for ${inbound.getPresenceAccounts().length} account(s)`);

src/channel.ts

@@ -21,6 +21,39 @@ import {
   resolveDefaultFabricAccountId,
   type ResolvedFabricAccount,
 } from './accounts.js';
+import { getChannelType } from './channel-meta.js';
+
+/**
+ * Map a Fabric channel xType to an openclaw routing peer.kind / ChatType.
+ *
+ * Fabric distinguishes channels by xType ('dm' | 'triage' | 'group' |
+ * 'broadcast' | 'announce' | ...). Openclaw's session router only knows
+ * 'direct' | 'group' | 'channel'. We collapse:
+ *   - 'dm'  → 'direct' (1:1 conversation; agent always speaks)
+ *   - rest  → 'group'  (multi-party; turn-engine gates speech)
+ *
+ * Sessions are keyed by peer.kind, so inbound and outbound MUST agree —
+ * otherwise the agent's outbound message lands in a different session
+ * than the inbound that triggered it and conversation state splits.
+ *
+ * Outbound has no live xType (the agent target is just a channelId), so
+ * it consults the channel-meta cache populated by inbound. Cache miss
+ * (channel never observed) falls back to 'group' — same as the pre-fix
+ * behavior, no regression on cold cache. The proactive-DM-first-message
+ * edge case (agent DMs a channel before any inbound) still lands as
+ * 'group' on that one outbound; the next inbound + outbound pair will
+ * agree on 'direct'.
+ */
+export type FabricPeerRouting = { peerKind: 'direct' | 'group'; chatType: 'direct' | 'group' };
+
+export function fabricPeerRoutingForXType(xType: string | null | undefined): FabricPeerRouting {
+  if (xType === 'dm') return { peerKind: 'direct', chatType: 'direct' };
+  return { peerKind: 'group', chatType: 'group' };
+}
+
+export function fabricPeerRoutingForChannel(channelId: string): FabricPeerRouting {
+  return fabricPeerRoutingForXType(getChannelType(channelId));
+}
 
 type AnyCfg = { channels?: { fabric?: unknown }; [k: string]: unknown };
 
@@ -45,13 +78,18 @@ export function looksLikeFabricTargetId(raw: string): boolean {
 export function resolveFabricOutboundSessionRoute(params: ChannelOutboundSessionRouteParams) {
   const id = stripFabricTargetPrefix(params.target);
   if (!id) return null;
+  // Consult the channel-meta cache populated by inbound — DM channels
+  // need peer.kind='direct' so the outbound session key matches the
+  // inbound one. Cache miss falls back to 'group' (the pre-fix default,
+  // no regression on cold cache).
+  const { peerKind, chatType } = fabricPeerRoutingForChannel(id);
   return buildChannelOutboundSessionRoute({
     cfg: params.cfg,
     agentId: params.agentId,
     channel: 'fabric',
     accountId: params.accountId,
-    peer: { kind: 'group', id },
+    peer: { kind: peerKind, id },
-    chatType: 'group',
+    chatType,
     from: `fabric:channel:${id}`,
     to: `fabric:${id}`,
   });

src/inbound.ts

@@ -6,6 +6,7 @@ import { dispatchInboundReplyWithBase } from 'openclaw/plugin-sdk/inbound-reply-
 import type { FabricClient, FabricSession } from './fabric-client.js';
 import type { IdentityRegistry } from './identity.js';
 import { resolveCoalesce } from './accounts.js';
+import { fabricPeerRoutingForXType } from './channel.js';
 import { recordChannelType } from './channel-meta.js';
 import { enqueueDelivery, flushFabricForChannel } from './coalesce.js';
 
@@ -280,17 +281,25 @@ export class FabricInbound {
     agentId: string;
     fabricUserId: string;
     guildBaseUrl: string;
+    guildNodeId: string;
     fabricApiKey: string;
   }> {
-    const out: Array<{ agentId: string; fabricUserId: string; guildBaseUrl: string; fabricApiKey: string }> = [];
+    const out: Array<{
+      agentId: string;
+      fabricUserId: string;
+      guildBaseUrl: string;
+      guildNodeId: string;
+      fabricApiKey: string;
+    }> = [];
     for (const entry of this.identity.list()) {
       if (!entry.fabricUserId) continue;
-      const presenceGuildUrl = this.firstGuildEndpointByAgent.get(entry.agentId);
+      const presenceGuild = this.firstGuildByAgent.get(entry.agentId);
-      if (!presenceGuildUrl) continue;
+      if (!presenceGuild) continue;
       out.push({
         agentId: entry.agentId,
         fabricUserId: entry.fabricUserId,
-        guildBaseUrl: presenceGuildUrl,
+        guildBaseUrl: presenceGuild.endpoint,
+        guildNodeId: presenceGuild.nodeId,
         fabricApiKey: entry.fabricApiKey,
       });
     }
@@ -298,24 +307,44 @@ export class FabricInbound {
   }
 
   // Filled by connectAgent for each (agent, guild). Tracks ONLY the first
-  // guild per agent (used as the presence-push target).
+  // guild per agent (used as the presence-push target). Stores both
-  private firstGuildEndpointByAgent = new Map<string, string>();
+  // endpoint and nodeId — presence-sync needs both: endpoint to build
+  // the URL, nodeId to pick the matching guildAccessToken from a fresh
+  // agent-login response.
+  private firstGuildByAgent = new Map<string, { endpoint: string; nodeId: string }>();
 
   private async connectAgent(agentId: string, session: FabricSession): Promise<void> {
     const selfUserId = session.user.id;
     // First-guild capture for presence-sync push target. session.guilds is
     // already in priority order from Center; we take the first one with a
     // valid endpoint and stop. Multi-guild presence is a future concern.
-    if (!this.firstGuildEndpointByAgent.has(agentId)) {
+    if (!this.firstGuildByAgent.has(agentId)) {
       const firstGuild = session.guilds.find((g) => typeof g.endpoint === 'string' && g.endpoint.length > 0);
-      if (firstGuild) this.firstGuildEndpointByAgent.set(agentId, firstGuild.endpoint);
+      if (firstGuild) this.firstGuildByAgent.set(agentId, { endpoint: firstGuild.endpoint, nodeId: firstGuild.nodeId });
     }
     for (const g of session.guilds) {
       const tok = session.guildAccessTokens.find((t) => t.guildNodeId === g.nodeId)?.token;
       if (!tok) continue;
+      // Use the *callback* form of `auth` so socket.io re-evaluates the JWT
+      // on every (re)connect. The single-shot `auth: { token: tok }` shape
+      // captured the token in closure: after socket.io's silent auto-reconnect
+      // the backend got the same JWT that expired ~15 min into the session
+      // (guildAccessToken TTL = 900s) and silently rejected the handshake at
+      // the application layer. The client's `connect` event still fired (TCP
+      // succeeded), so the plugin happily ran the channel-resync, emitted
+      // `join_channel` into the void, and logged "joined N channel(s)" while
+      // the backend was actually broadcasting message.created to a room with
+      // zero subscribers. End user symptom: DMs to agents silently dropped.
       const socket = io(`${g.endpoint}/realtime`, {
         transports: ['websocket'],
-        auth: { token: tok },
+        auth: (cb) => {
+          // Best-effort fresh token; on transient failure fall back to the
+          // last known good one. tokenCache also keeps HTTP calls (attachment
+          // download / reply post) from 401'ing in the same window.
+          this.freshGuildToken(agentId, g.nodeId, session)
+            .then((fresh) => cb({ token: fresh ?? tok }))
+            .catch(() => cb({ token: tok }));
+        },
         autoConnect: false,
       });
       // Tracked socket.io rooms for this (agent, guild). The initial fetch
@@ -504,11 +533,19 @@ export class FabricInbound {
     const core = this.core as Core & Record<string, unknown>;
     const cfg = this.cfg as { session?: { store?: unknown } };
     try {
+      // Route by xType. DM channels need peer.kind='direct' so openclaw
+      // treats them as 1:1 (sessionKey 'agent:<id>:fabric:direct:<chan>'
+      // and ctx.ChatType='direct') rather than as a multi-party group.
+      // Without this, the agent's user-prompt metadata says
+      // 'is_group_chat: true' on a DM and downstream prompt logic
+      // (commands-handlers `isDirectMessage` checks ChatType==='direct')
+      // misclassifies the turn.
+      const { peerKind, chatType } = fabricPeerRoutingForXType(m.xType);
       const route = core.channel.routing.resolveAgentRoute({
         cfg: this.cfg,
         channel: 'fabric',
         accountId: agentId,
-        peer: { kind: 'group', id: channelId },
+        peer: { kind: peerKind, id: channelId },
       });
       const storePath = core.channel.session.resolveStorePath(cfg.session?.store, {
         agentId: route.agentId,
@@ -523,7 +560,7 @@ export class FabricInbound {
         To: `fabric:${channelId}`,
         SessionKey: route.sessionKey,
         AccountId: route.accountId ?? agentId,
-        ChatType: 'group',
+        ChatType: chatType,
         ConversationLabel: `fabric:${guild.nodeId}`,
         SenderId: m.authorUserId ?? 'fabric',
         Provider: 'fabric',

src/presence-sync.ts

@@ -2,18 +2,26 @@
  * presence-sync — read each connected agent's HF status (via the
  * cross-plugin `globalThis.__hfAgentStatus.get(agentId)` exposed by
  * HarborForge.OpenclawPlugin) and push diffs to Fabric.Backend.Guild
- * `PUT /agents/:userId/presence` so the backend can apply busy-discard
+ * `PUT /api/agents/:userId/presence` so the backend can apply
- * on `announce`-type channel deliveries.
+ * busy-discard on `announce`-type channel deliveries.
  *
  * Push model: we only PUT when an agent's status actually changes
  * (since the last push). The HF-side accessor has its own TTL cache
  * to absorb the every-30s polling.
  *
+ * Auth: the endpoint sits behind ApiKeyGuard (global APP_GUARD per
+ * app.module.js) which expects `Authorization: Bearer <guild-token>`
+ * — NOT the agent's fabricApiKey directly. So before each PUT we do
+ * a fresh agent-login (or reuse a cached token if still within its
+ * 15-min JWT TTL) and pull the guildAccessToken matching the target
+ * guild. Status changes are rare enough that login overhead is fine.
+ *
  * If HF plugin isn't loaded (`__hfAgentStatus` undefined), the loop
  * is a no-op — Fabric backend defaults presence to 'unknown' which is
  * treated as not-busy. Announce-channel delivery still works; busy
  * filtering simply doesn't kick in.
  */
+import type { FabricClient } from './fabric-client.js';
 
 type HfStatus = 'idle' | 'on_call' | 'busy' | 'exhausted' | 'offline';
 type Bridge = { get(agentId: string): Promise<HfStatus | undefined> };
@@ -23,15 +31,36 @@ export interface PresenceSyncAccount {
   agentId: string;
   fabricUserId: string; // the agent's Fabric Center user id (UUID)
   guildBaseUrl: string; // e.g. https://fabric.hangman-lab.top/guild/<id>
-  fabricApiKey: string; // existing per-account key
+  guildNodeId: string;  // which guildAccessTokens[].guildNodeId to pick
+  fabricApiKey: string; // existing per-account key (used for agent-login)
+}
+
+// Guild access JWTs expire every 900s. Refresh ~2 min early to stay
+// safely inside the window even if a tick runs late.
+const TOKEN_TTL_MS = (15 - 2) * 60 * 1000;
+
+interface CachedToken {
+  token: string;
+  expiresAt: number; // epoch ms
 }
 
 export class PresenceSync {
   private timer: ReturnType<typeof setInterval> | null = null;
   private readonly lastStatus = new Map<string, HfStatus>(); // by agentId
   private readonly accounts = new Map<string, PresenceSyncAccount>();
+  private readonly tokenCache = new Map<string, CachedToken>(); // by agentId
 
-  constructor(private readonly logger: Logger) {}
+  // Mutex flag: a tick iterates accounts serially with `await` on each
+  // agent-login + PUT round-trip, so a single tick can easily run 20+s
+  // when there are many accounts. setInterval(intervalMs) does NOT wait
+  // for the previous tick to finish — without this guard the next tick
+  // fires on top of a still-running one and two parallel iterations
+  // PUT the same agentId within milliseconds. That tipped the backend's
+  // first-time-insert race (separate fix in Fabric.Backend.Guild) into
+  // 500s on prod. Guarded ticks just skip a beat instead.
+  private inflight = false;
+
+  constructor(private readonly logger: Logger, private readonly client: FabricClient) {}
 
   setAccounts(accounts: PresenceSyncAccount[]): void {
     this.accounts.clear();
@@ -54,7 +83,50 @@ export class PresenceSync {
     }
   }
 
+  /**
+   * Fetch a fresh guildAccessToken for `acct`, caching it under the
+   * agentId until just before its JWT expiry. Returns null on login
+   * failure or if the session has no matching guild — caller logs +
+   * skips the PUT.
+   */
+  private async ensureGuildToken(acct: PresenceSyncAccount): Promise<string | null> {
+    const now = Date.now();
+    const cached = this.tokenCache.get(acct.agentId);
+    if (cached && cached.expiresAt > now) return cached.token;
+
+    let session;
+    try {
+      session = await this.client.agentLogin(acct.fabricApiKey);
+    } catch (err) {
+      this.logger.warn(`fabric: presence-sync agent-login failed for ${acct.agentId}: ${String(err)}`);
+      return null;
+    }
+    const entry = session.guildAccessTokens.find((g) => g.guildNodeId === acct.guildNodeId);
+    if (!entry?.token) {
+      this.logger.warn(
+        `fabric: presence-sync no guild token for ${acct.agentId} guild=${acct.guildNodeId}`,
+      );
+      return null;
+    }
+    this.tokenCache.set(acct.agentId, { token: entry.token, expiresAt: now + TOKEN_TTL_MS });
+    return entry.token;
+  }
+
   private async tick(): Promise<void> {
+    // Mutex: see the `inflight` field declaration for the why. Drop
+    // overlapping ticks rather than letting them run concurrently —
+    // status is gated by `lastStatus !== bridge.get`, so skipping a
+    // beat costs nothing the next beat won't catch.
+    if (this.inflight) return;
+    this.inflight = true;
+    try {
+      await this.tickInner();
+    } finally {
+      this.inflight = false;
+    }
+  }
+
+  private async tickInner(): Promise<void> {
     const bridge = (globalThis as Record<string, unknown>)['__hfAgentStatus'] as Bridge | undefined;
     if (!bridge || typeof bridge.get !== 'function') return; // HF plugin not loaded — skip
 
@@ -68,13 +140,22 @@ export class PresenceSync {
       if (!status) continue;
       if (this.lastStatus.get(agentId) === status) continue; // no change → no PUT
 
+      const guildToken = await this.ensureGuildToken(acct);
+      if (!guildToken) continue;
+
       try {
-        const url = `${acct.guildBaseUrl.replace(/\/$/, '')}/agents/${encodeURIComponent(acct.fabricUserId)}/presence`;
+        // Endpoint: PUT /api/agents/:userId/presence. ApiKeyGuard (global
+        // APP_GUARD) requires `Authorization: Bearer <guildAccessToken>`
+        // — NOT the agent's raw fabricApiKey. Pre-v1: this loop sent
+        // x-api-key and got 401 "missing bearer token" forever. The /api
+        // prefix is required because the guild backend sets a global
+        // 'api' prefix in main.ts setGlobalPrefix('api').
+        const url = `${acct.guildBaseUrl.replace(/\/$/, '')}/api/agents/${encodeURIComponent(acct.fabricUserId)}/presence`;
         const res = await fetch(url, {
           method: 'PUT',
           headers: {
             'content-type': 'application/json',
-            'x-api-key': acct.fabricApiKey,
+            authorization: `Bearer ${guildToken}`,
           },
           body: JSON.stringify({ status, source: 'hf-plugin' }),
         });
@@ -82,6 +163,10 @@ export class PresenceSync {
           this.lastStatus.set(agentId, status);
           this.logger.info(`fabric: presence-sync ${agentId} → ${status}`);
         } else {
+          // 401 here usually means the cached token went stale unexpectedly
+          // (server-side rotation or clock skew) — drop the cache so the
+          // next tick re-logs-in.
+          if (res.status === 401) this.tokenCache.delete(agentId);
           this.logger.warn(`fabric: presence-sync PUT ${agentId} failed: ${res.status}`);
         }
       } catch (err) {