fix: honor apiKey auth-choice tokens (#1485) (thanks @iHildy)

Adds sanitization to extractAssistantText in sessions-helpers.ts to
prevent tool call text from leaking to users. Previously, messages
retrieved from chat history via sessions-helpers.ts could expose:

- Minimax XML tool calls (<invoke>...</invoke>)
- Downgraded tool call markers ([Tool Call: name (ID: ...)])
- Thinking tags (<think>...</think>)

This fix:
- Exports the stripping functions from pi-embedded-utils.ts
- Adds a new sanitizeTextContent helper in sessions-helpers.ts
- Updates extractAssistantText to sanitize before returning
- Updates extractMessageText in commands-subagents.ts to sanitize

Fixes #1269

Co-Authored-By: Claude <noreply@anthropic.com>

CHANGELOG.md

@@ -6,6 +6,7 @@ Docs: https://docs.clawd.bot
 
 ### Changes
 - Highlight: Lobster optional plugin tool for typed workflows + approval gates. https://docs.clawd.bot/tools/lobster
+- Highlight: Compaction safeguard now uses adaptive chunking, progressive fallback, and UI status + retries. (#1466) Thanks @dlauer.
 - Lobster: allow workflow file args via `argsJson` in the plugin tool. https://docs.clawd.bot/tools/lobster
 - Agents: add identity avatar config support and Control UI avatar rendering. (#1329, #1424) Thanks @dlauer.
 - Slack: add chat-type reply threading overrides via `replyToModeByChatType`. (#1442) Thanks @stefangalescu.
@@ -28,9 +29,12 @@ Docs: https://docs.clawd.bot
 - BlueBubbles: stop typing indicator on idle/no-reply. (#1439) Thanks @Nicell.
 - Message tool: keep path/filePath as-is for send; hydrate buffers only for sendAttachment. (#1444) Thanks @hopyky.
 - Auto-reply: only report a model switch when session state is available. (#1465) Thanks @robbyczgw-cla.
-- Compaction: harden oversized summarization handling and gate UI compaction status by verbose level. (#1466) Thanks @dlauer.
 - Control UI: resolve local avatar URLs with basePath across injection + identity RPC. (#1457) Thanks @dlauer.
+- Agents: sanitize assistant history text to strip tool-call markers. (#1456) Thanks @zerone0x.
+- Discord: clarify Message Content Intent onboarding hint. (#1487) Thanks @kyleok.
+- Onboarding: honor `--auth-choice apiKey` token provider + token flags to skip prompts. (#1485) Thanks @iHildy.
 - Agents: surface concrete API error details instead of generic AI service errors.
+- Exec: fall back to non-PTY when PTY spawn fails (EBADF). (#1484)
 - Exec approvals: allow per-segment allowlists for chained shell commands on gateway + node hosts. (#1458) Thanks @czekaj.
 - Agents: make OpenAI sessions image-sanitize-only; gate tool-id/repair sanitization by provider.
 - Doctor: honor CLAWDBOT_GATEWAY_TOKEN for auth checks and security audit token reuse. (#1448) Thanks @azade-c.
@@ -43,8 +47,10 @@ Docs: https://docs.clawd.bot
 - Auth: skip auth profiles in cooldown during initial selection and rotation. (#1316) Thanks @odrobnik.
 - Agents/TUI: honor user-pinned auth profiles during cooldown and preserve search picker ranking. (#1432) Thanks @tobiasbischoff.
 - Docs: fix gog auth services example to include docs scope. (#1454) Thanks @zerone0x.
+- Slack: reduce WebClient retries to avoid duplicate sends. (#1481)
 - Slack: read thread replies for message reads when threadId is provided (replies-only). (#1450) Thanks @rodrigouroz.
 - macOS: prefer linked channels in gateway summary to avoid false “not linked” status.
+- macOS/tests: fix gateway summary lookup after guard unwrap; prevent browser opens during tests. (ECID-1483)
 - Providers: improve GitHub Copilot integration (enterprise support, base URL, and auth flow alignment).
 
 ## 2026.1.21-2

apps/macos/Sources/Clawdbot/DebugSettings.swift

@@ -102,7 +102,9 @@ struct DebugSettings: View {
                         }
                     }
 
-                Text("When enabled, Clawdbot won't install or manage \(gatewayLaunchdLabel). It will only attach to an existing Gateway.")
+                Text(
+                    "When enabled, Clawdbot won't install or manage \(gatewayLaunchdLabel). " +
+                        "It will only attach to an existing Gateway.")
                     .font(.caption)
                     .foregroundStyle(.secondary)
 

apps/macos/Sources/Clawdbot/GatewayProcessManager.swift

@@ -248,12 +248,11 @@ final class GatewayProcessManager {
             guard let linkId else {
                 return "port \(port), health probe succeeded, \(instanceText)"
             }
-            let linked = linkId.flatMap { snap.channels[$0]?.linked } ?? false
-            let authAge = linkId.flatMap { snap.channels[$0]?.authAgeMs }.flatMap(msToAge) ?? "unknown age"
+            let linked = snap.channels[linkId]?.linked ?? false
+            let authAge = snap.channels[linkId]?.authAgeMs.flatMap(msToAge) ?? "unknown age"
             let label =
-                linkId.flatMap { snap.channelLabels?[$0] } ??
-                linkId?.capitalized ??
-                "channel"
+                snap.channelLabels?[linkId] ??
+                linkId.capitalized
             let linkText = linked ? "linked" : "not linked"
             return "port \(port), \(label) \(linkText), auth \(authAge), \(instanceText)"
         }

docs/cli/index.md

@@ -295,8 +295,8 @@ Options:
 - `--mode <local|remote>`
 - `--flow <quickstart|advanced|manual>` (manual is an alias for advanced)
 - `--auth-choice <setup-token|claude-cli|token|openai-codex|openai-api-key|openrouter-api-key|ai-gateway-api-key|moonshot-api-key|kimi-code-api-key|codex-cli|gemini-api-key|zai-api-key|apiKey|minimax-api|opencode-zen|skip>`
-- `--token-provider <id>` (non-interactive; used with `--auth-choice token`)
-- `--token <token>` (non-interactive; used with `--auth-choice token`)
+- `--token-provider <id>` (non-interactive; used with `--auth-choice token` or `apiKey`)
+- `--token <token>` (non-interactive; used with `--auth-choice token` or `apiKey`)
 - `--token-profile-id <id>` (non-interactive; default: `<provider>:manual`)
 - `--token-expires-in <duration>` (non-interactive; e.g. `365d`, `12h`)
 - `--anthropic-api-key <key>`

extensions/memory-lancedb/index.test.ts

@@ -14,11 +14,12 @@ import fs from "node:fs/promises";
 import path from "node:path";
 import os from "node:os";
 
-// Skip if no OpenAI API key
-const OPENAI_API_KEY = process.env.OPENAI_API_KEY;
-const describeWithKey = OPENAI_API_KEY ? describe : describe.skip;
+const OPENAI_API_KEY = process.env.OPENAI_API_KEY ?? "test-key";
+const HAS_OPENAI_KEY = Boolean(process.env.OPENAI_API_KEY);
+const liveEnabled = HAS_OPENAI_KEY && process.env.CLAWDBOT_LIVE_TEST === "1";
+const describeLive = liveEnabled ? describe : describe.skip;
 
-describeWithKey("memory plugin e2e", () => {
+describe("memory plugin e2e", () => {
   let tmpDir: string;
   let dbPath: string;
 
@@ -159,7 +160,7 @@ describeWithKey("memory plugin e2e", () => {
 });
 
 // Live tests that require OpenAI API key and actually use LanceDB
-describeWithKey("memory plugin live tests", () => {
+describeLive("memory plugin live tests", () => {
   let tmpDir: string;
   let dbPath: string;
 
@@ -176,6 +177,7 @@ describeWithKey("memory plugin live tests", () => {
 
   test("memory tools work end-to-end", async () => {
     const { default: memoryPlugin } = await import("./index.js");
+    const liveApiKey = process.env.OPENAI_API_KEY ?? "";
 
     // Mock plugin API
     const registeredTools: any[] = [];
@@ -191,7 +193,7 @@ describeWithKey("memory plugin live tests", () => {
       config: {},
       pluginConfig: {
         embedding: {
-          apiKey: OPENAI_API_KEY,
+          apiKey: liveApiKey,
           model: "text-embedding-3-small",
         },
         dbPath,

src/agents/bash-tools.exec.pty-fallback.test.ts

@@ -0,0 +1,31 @@
+import { afterEach, expect, test, vi } from "vitest";
+
+import { resetProcessRegistryForTests } from "./bash-process-registry";
+
+afterEach(() => {
+  resetProcessRegistryForTests();
+  vi.resetModules();
+  vi.clearAllMocks();
+});
+
+test("exec falls back when PTY spawn fails", async () => {
+  vi.doMock("@lydell/node-pty", () => ({
+    spawn: () => {
+      const err = new Error("spawn EBADF");
+      (err as NodeJS.ErrnoException).code = "EBADF";
+      throw err;
+    },
+  }));
+
+  const { createExecTool } = await import("./bash-tools.exec");
+  const tool = createExecTool({ allowBackground: false });
+  const result = await tool.execute("toolcall", {
+    command: "printf ok",
+    pty: true,
+  });
+
+  expect(result.details.status).toBe("completed");
+  const text = result.content?.[0]?.text ?? "";
+  expect(text).toContain("ok");
+  expect(text).toContain("PTY spawn failed");
+});

src/agents/bash-tools.exec.ts

@@ -26,7 +26,7 @@ import {
   resolveShellEnvFallbackTimeoutMs,
 } from "../infra/shell-env.js";
 import { enqueueSystemEvent } from "../infra/system-events.js";
-import { logInfo } from "../logger.js";
+import { logInfo, logWarn } from "../logger.js";
 import {
   type ProcessSession,
   type SessionStdin,
@@ -381,41 +381,56 @@ async function runExecProcess(opts: {
     ) as ChildProcessWithoutNullStreams;
     stdin = child.stdin;
   } else if (opts.usePty) {
-    const ptyModule = (await import("@lydell/node-pty")) as unknown as {
-      spawn?: PtySpawn;
-      default?: { spawn?: PtySpawn };
-    };
-    const spawnPty = ptyModule.spawn ?? ptyModule.default?.spawn;
-    if (!spawnPty) {
-      throw new Error("PTY support is unavailable (node-pty spawn not found).");
-    }
     const { shell, args: shellArgs } = getShellConfig();
-    pty = spawnPty(shell, [...shellArgs, opts.command], {
-      cwd: opts.workdir,
-      env: opts.env,
-      name: process.env.TERM ?? "xterm-256color",
-      cols: 120,
-      rows: 30,
-    });
-    stdin = {
-      destroyed: false,
-      write: (data, cb) => {
-        try {
-          pty?.write(data);
-          cb?.(null);
-        } catch (err) {
-          cb?.(err as Error);
-        }
-      },
-      end: () => {
-        try {
-          const eof = process.platform === "win32" ? "\x1a" : "\x04";
-          pty?.write(eof);
-        } catch {
-          // ignore EOF errors
-        }
-      },
-    };
+    try {
+      const ptyModule = (await import("@lydell/node-pty")) as unknown as {
+        spawn?: PtySpawn;
+        default?: { spawn?: PtySpawn };
+      };
+      const spawnPty = ptyModule.spawn ?? ptyModule.default?.spawn;
+      if (!spawnPty) {
+        throw new Error("PTY support is unavailable (node-pty spawn not found).");
+      }
+      pty = spawnPty(shell, [...shellArgs, opts.command], {
+        cwd: opts.workdir,
+        env: opts.env,
+        name: process.env.TERM ?? "xterm-256color",
+        cols: 120,
+        rows: 30,
+      });
+      stdin = {
+        destroyed: false,
+        write: (data, cb) => {
+          try {
+            pty?.write(data);
+            cb?.(null);
+          } catch (err) {
+            cb?.(err as Error);
+          }
+        },
+        end: () => {
+          try {
+            const eof = process.platform === "win32" ? "\x1a" : "\x04";
+            pty?.write(eof);
+          } catch {
+            // ignore EOF errors
+          }
+        },
+      };
+    } catch (err) {
+      const errText = String(err);
+      const warning = `Warning: PTY spawn failed (${errText}); retrying without PTY for \`${opts.command}\`.`;
+      logWarn(`exec: PTY spawn failed (${errText}); retrying without PTY for "${opts.command}".`);
+      opts.warnings.push(warning);
+      child = spawn(shell, [...shellArgs, opts.command], {
+        cwd: opts.workdir,
+        env: opts.env,
+        detached: process.platform !== "win32",
+        stdio: ["pipe", "pipe", "pipe"],
+        windowsHide: true,
+      }) as ChildProcessWithoutNullStreams;
+      stdin = child.stdin;
+    }
   } else {
     const { shell, args: shellArgs } = getShellConfig();
     child = spawn(shell, [...shellArgs, opts.command], {
@@ -1320,7 +1335,7 @@ export function createExecTool(
 
       const effectiveTimeout =
         typeof params.timeout === "number" ? params.timeout : defaultTimeoutSec;
-      const warningText = warnings.length ? `${warnings.join("\n")}\n\n` : "";
+      const getWarningText = () => (warnings.length ? `${warnings.join("\n")}\n\n` : "");
       const usePty = params.pty === true && !sandbox;
       const run = await runExecProcess({
         command: params.command,
@@ -1360,7 +1375,7 @@ export function createExecTool(
               {
                 type: "text",
                 text:
-                  `${warningText}` +
+                  `${getWarningText()}` +
                   `Command still running (session ${run.session.id}, pid ${
                     run.session.pid ?? "n/a"
                   }). ` +
@@ -1410,7 +1425,7 @@ export function createExecTool(
               content: [
                 {
                   type: "text",
-                  text: `${warningText}${outcome.aggregated || "(no output)"}`,
+                  text: `${getWarningText()}${outcome.aggregated || "(no output)"}`,
                 },
               ],
               details: {

src/agents/clawdbot-tools.subagents.sessions-spawn-announces-back-requester-group-channel.test.ts

@@ -1,235 +0,0 @@
-import { beforeEach, describe, expect, it, vi } from "vitest";
-
-const callGatewayMock = vi.fn();
-vi.mock("../gateway/call.js", () => ({
-  callGateway: (opts: unknown) => callGatewayMock(opts),
-}));
-
-let configOverride: ReturnType<(typeof import("../config/config.js"))["loadConfig"]> = {
-  session: {
-    mainKey: "main",
-    scope: "per-sender",
-  },
-};
-
-vi.mock("../config/config.js", async (importOriginal) => {
-  const actual = await importOriginal<typeof import("../config/config.js")>();
-  return {
-    ...actual,
-    loadConfig: () => configOverride,
-    resolveGatewayPort: () => 18789,
-  };
-});
-
-import { emitAgentEvent } from "../infra/agent-events.js";
-import { createClawdbotTools } from "./clawdbot-tools.js";
-import { resetSubagentRegistryForTests } from "./subagent-registry.js";
-
-describe("clawdbot-tools: subagents", () => {
-  beforeEach(() => {
-    configOverride = {
-      session: {
-        mainKey: "main",
-        scope: "per-sender",
-      },
-    };
-  });
-
-  it("sessions_spawn runs cleanup via lifecycle events", async () => {
-    resetSubagentRegistryForTests();
-    callGatewayMock.mockReset();
-    const calls: Array<{ method?: string; params?: unknown }> = [];
-    let agentCallCount = 0;
-    let deletedKey: string | undefined;
-    let childRunId: string | undefined;
-    let childSessionKey: string | undefined;
-    const waitCalls: Array<{ runId?: string; timeoutMs?: number }> = [];
-
-    callGatewayMock.mockImplementation(async (opts: unknown) => {
-      const request = opts as { method?: string; params?: unknown };
-      calls.push(request);
-      if (request.method === "agent") {
-        agentCallCount += 1;
-        const runId = `run-${agentCallCount}`;
-        const params = request.params as {
-          message?: string;
-          sessionKey?: string;
-          channel?: string;
-          timeout?: number;
-          lane?: string;
-        };
-        // Only capture the first agent call (subagent spawn, not main agent trigger)
-        if (params?.lane === "subagent") {
-          childRunId = runId;
-          childSessionKey = params?.sessionKey ?? "";
-          expect(params?.channel).toBe("discord");
-          expect(params?.timeout).toBe(1);
-        }
-        return {
-          runId,
-          status: "accepted",
-          acceptedAt: 1000 + agentCallCount,
-        };
-      }
-      if (request.method === "agent.wait") {
-        const params = request.params as { runId?: string; timeoutMs?: number } | undefined;
-        waitCalls.push(params ?? {});
-        // Return "ok" with timing info for the child run
-        return { runId: params?.runId ?? "run-1", status: "ok", startedAt: 1000, endedAt: 2000 };
-      }
-      if (request.method === "sessions.delete") {
-        const params = request.params as { key?: string } | undefined;
-        deletedKey = params?.key;
-        return { ok: true };
-      }
-      return {};
-    });
-
-    const tool = createClawdbotTools({
-      agentSessionKey: "discord:group:req",
-      agentChannel: "discord",
-    }).find((candidate) => candidate.name === "sessions_spawn");
-    if (!tool) throw new Error("missing sessions_spawn tool");
-
-    const result = await tool.execute("call1", {
-      task: "do thing",
-      runTimeoutSeconds: 1,
-      cleanup: "delete",
-    });
-    expect(result.details).toMatchObject({
-      status: "accepted",
-      runId: "run-1",
-    });
-
-    if (!childRunId) throw new Error("missing child runId");
-    emitAgentEvent({
-      runId: childRunId,
-      stream: "lifecycle",
-      data: {
-        phase: "end",
-        startedAt: 1234,
-        endedAt: 2345,
-      },
-    });
-
-    await new Promise((resolve) => setTimeout(resolve, 0));
-    await new Promise((resolve) => setTimeout(resolve, 0));
-    await new Promise((resolve) => setTimeout(resolve, 0));
-
-    const childWait = waitCalls.find((call) => call.runId === childRunId);
-    expect(childWait?.timeoutMs).toBe(1000);
-
-    // Two agent calls: subagent spawn + main agent trigger
-    const agentCalls = calls.filter((call) => call.method === "agent");
-    expect(agentCalls).toHaveLength(2);
-
-    // First call: subagent spawn
-    const first = agentCalls[0]?.params as
-      | {
-          lane?: string;
-          deliver?: boolean;
-          sessionKey?: string;
-          channel?: string;
-        }
-      | undefined;
-    expect(first?.lane).toBe("subagent");
-    expect(first?.deliver).toBe(false);
-    expect(first?.channel).toBe("discord");
-    expect(first?.sessionKey?.startsWith("agent:main:subagent:")).toBe(true);
-    expect(childSessionKey?.startsWith("agent:main:subagent:")).toBe(true);
-
-    // Second call: main agent trigger with announce message
-    const second = agentCalls[1]?.params as
-      | {
-          sessionKey?: string;
-          message?: string;
-          deliver?: boolean;
-        }
-      | undefined;
-    expect(second?.sessionKey).toBe("discord:group:req");
-    expect(second?.deliver).toBe(true);
-    expect(second?.message).toContain("background task");
-
-    // No direct send to external channel (main agent handles delivery)
-    const sendCalls = calls.filter((c) => c.method === "send");
-    expect(sendCalls.length).toBe(0);
-
-    // Session should be deleted since cleanup=delete
-    expect(deletedKey?.startsWith("agent:main:subagent:")).toBe(true);
-  });
-
-  it("sessions_spawn announces with requester accountId", async () => {
-    resetSubagentRegistryForTests();
-    callGatewayMock.mockReset();
-    const calls: Array<{ method?: string; params?: unknown }> = [];
-    let agentCallCount = 0;
-    let childRunId: string | undefined;
-
-    callGatewayMock.mockImplementation(async (opts: unknown) => {
-      const request = opts as { method?: string; params?: unknown };
-      calls.push(request);
-      if (request.method === "agent") {
-        agentCallCount += 1;
-        const runId = `run-${agentCallCount}`;
-        const params = request.params as { lane?: string; sessionKey?: string } | undefined;
-        if (params?.lane === "subagent") {
-          childRunId = runId;
-        }
-        return {
-          runId,
-          status: "accepted",
-          acceptedAt: 4000 + agentCallCount,
-        };
-      }
-      if (request.method === "agent.wait") {
-        const params = request.params as { runId?: string; timeoutMs?: number } | undefined;
-        return { runId: params?.runId ?? "run-1", status: "ok", startedAt: 1000, endedAt: 2000 };
-      }
-      if (request.method === "sessions.delete" || request.method === "sessions.patch") {
-        return { ok: true };
-      }
-      return {};
-    });
-
-    const tool = createClawdbotTools({
-      agentSessionKey: "main",
-      agentChannel: "whatsapp",
-      agentAccountId: "kev",
-    }).find((candidate) => candidate.name === "sessions_spawn");
-    if (!tool) throw new Error("missing sessions_spawn tool");
-
-    const result = await tool.execute("call2", {
-      task: "do thing",
-      runTimeoutSeconds: 1,
-      cleanup: "keep",
-    });
-    expect(result.details).toMatchObject({
-      status: "accepted",
-      runId: "run-1",
-    });
-
-    if (!childRunId) throw new Error("missing child runId");
-    emitAgentEvent({
-      runId: childRunId,
-      stream: "lifecycle",
-      data: {
-        phase: "end",
-        startedAt: 1000,
-        endedAt: 2000,
-      },
-    });
-
-    await new Promise((resolve) => setTimeout(resolve, 0));
-    await new Promise((resolve) => setTimeout(resolve, 0));
-    await new Promise((resolve) => setTimeout(resolve, 0));
-
-    const agentCalls = calls.filter((call) => call.method === "agent");
-    expect(agentCalls).toHaveLength(2);
-    const announceParams = agentCalls[1]?.params as
-      | { accountId?: string; channel?: string; deliver?: boolean }
-      | undefined;
-    expect(announceParams?.deliver).toBe(true);
-    expect(announceParams?.channel).toBe("whatsapp");
-    expect(announceParams?.accountId).toBe("kev");
-  });
-});

src/agents/clawdbot-tools.subagents.sessions-spawn-normalizes-allowlisted-agent-ids.test.ts

@@ -21,6 +21,7 @@ vi.mock("../config/config.js", async (importOriginal) => {
   };
 });
 
+import { emitAgentEvent } from "../infra/agent-events.js";
 import { createClawdbotTools } from "./clawdbot-tools.js";
 import { resetSubagentRegistryForTests } from "./subagent-registry.js";
 
@@ -120,4 +121,195 @@ describe("clawdbot-tools: subagents", () => {
     });
     expect(callGatewayMock).not.toHaveBeenCalled();
   });
+
+  it("sessions_spawn runs cleanup via lifecycle events", async () => {
+    resetSubagentRegistryForTests();
+    callGatewayMock.mockReset();
+    const calls: Array<{ method?: string; params?: unknown }> = [];
+    let agentCallCount = 0;
+    let deletedKey: string | undefined;
+    let childRunId: string | undefined;
+    let childSessionKey: string | undefined;
+    const waitCalls: Array<{ runId?: string; timeoutMs?: number }> = [];
+
+    callGatewayMock.mockImplementation(async (opts: unknown) => {
+      const request = opts as { method?: string; params?: unknown };
+      calls.push(request);
+      if (request.method === "agent") {
+        agentCallCount += 1;
+        const runId = `run-${agentCallCount}`;
+        const params = request.params as {
+          message?: string;
+          sessionKey?: string;
+          channel?: string;
+          timeout?: number;
+          lane?: string;
+        };
+        if (params?.lane === "subagent") {
+          childRunId = runId;
+          childSessionKey = params?.sessionKey ?? "";
+          expect(params?.channel).toBe("discord");
+          expect(params?.timeout).toBe(1);
+        }
+        return {
+          runId,
+          status: "accepted",
+          acceptedAt: 1000 + agentCallCount,
+        };
+      }
+      if (request.method === "agent.wait") {
+        const params = request.params as { runId?: string; timeoutMs?: number } | undefined;
+        waitCalls.push(params ?? {});
+        return { runId: params?.runId ?? "run-1", status: "ok", startedAt: 1000, endedAt: 2000 };
+      }
+      if (request.method === "sessions.delete") {
+        const params = request.params as { key?: string } | undefined;
+        deletedKey = params?.key;
+        return { ok: true };
+      }
+      return {};
+    });
+
+    const tool = createClawdbotTools({
+      agentSessionKey: "discord:group:req",
+      agentChannel: "discord",
+    }).find((candidate) => candidate.name === "sessions_spawn");
+    if (!tool) throw new Error("missing sessions_spawn tool");
+
+    const result = await tool.execute("call1", {
+      task: "do thing",
+      runTimeoutSeconds: 1,
+      cleanup: "delete",
+    });
+    expect(result.details).toMatchObject({
+      status: "accepted",
+      runId: "run-1",
+    });
+
+    if (!childRunId) throw new Error("missing child runId");
+    emitAgentEvent({
+      runId: childRunId,
+      stream: "lifecycle",
+      data: {
+        phase: "end",
+        startedAt: 1234,
+        endedAt: 2345,
+      },
+    });
+
+    await new Promise((resolve) => setTimeout(resolve, 0));
+    await new Promise((resolve) => setTimeout(resolve, 0));
+    await new Promise((resolve) => setTimeout(resolve, 0));
+
+    const childWait = waitCalls.find((call) => call.runId === childRunId);
+    expect(childWait?.timeoutMs).toBe(1000);
+
+    const agentCalls = calls.filter((call) => call.method === "agent");
+    expect(agentCalls).toHaveLength(2);
+
+    const first = agentCalls[0]?.params as
+      | {
+          lane?: string;
+          deliver?: boolean;
+          sessionKey?: string;
+          channel?: string;
+        }
+      | undefined;
+    expect(first?.lane).toBe("subagent");
+    expect(first?.deliver).toBe(false);
+    expect(first?.channel).toBe("discord");
+    expect(first?.sessionKey?.startsWith("agent:main:subagent:")).toBe(true);
+    expect(childSessionKey?.startsWith("agent:main:subagent:")).toBe(true);
+
+    const second = agentCalls[1]?.params as
+      | {
+          sessionKey?: string;
+          message?: string;
+          deliver?: boolean;
+        }
+      | undefined;
+    expect(second?.sessionKey).toBe("discord:group:req");
+    expect(second?.deliver).toBe(true);
+    expect(second?.message).toContain("background task");
+
+    const sendCalls = calls.filter((c) => c.method === "send");
+    expect(sendCalls.length).toBe(0);
+
+    expect(deletedKey?.startsWith("agent:main:subagent:")).toBe(true);
+  });
+
+  it("sessions_spawn announces with requester accountId", async () => {
+    resetSubagentRegistryForTests();
+    callGatewayMock.mockReset();
+    const calls: Array<{ method?: string; params?: unknown }> = [];
+    let agentCallCount = 0;
+    let childRunId: string | undefined;
+
+    callGatewayMock.mockImplementation(async (opts: unknown) => {
+      const request = opts as { method?: string; params?: unknown };
+      calls.push(request);
+      if (request.method === "agent") {
+        agentCallCount += 1;
+        const runId = `run-${agentCallCount}`;
+        const params = request.params as { lane?: string; sessionKey?: string } | undefined;
+        if (params?.lane === "subagent") {
+          childRunId = runId;
+        }
+        return {
+          runId,
+          status: "accepted",
+          acceptedAt: 4000 + agentCallCount,
+        };
+      }
+      if (request.method === "agent.wait") {
+        const params = request.params as { runId?: string; timeoutMs?: number } | undefined;
+        return { runId: params?.runId ?? "run-1", status: "ok", startedAt: 1000, endedAt: 2000 };
+      }
+      if (request.method === "sessions.delete" || request.method === "sessions.patch") {
+        return { ok: true };
+      }
+      return {};
+    });
+
+    const tool = createClawdbotTools({
+      agentSessionKey: "main",
+      agentChannel: "whatsapp",
+      agentAccountId: "kev",
+    }).find((candidate) => candidate.name === "sessions_spawn");
+    if (!tool) throw new Error("missing sessions_spawn tool");
+
+    const result = await tool.execute("call2", {
+      task: "do thing",
+      runTimeoutSeconds: 1,
+      cleanup: "keep",
+    });
+    expect(result.details).toMatchObject({
+      status: "accepted",
+      runId: "run-1",
+    });
+
+    if (!childRunId) throw new Error("missing child runId");
+    emitAgentEvent({
+      runId: childRunId,
+      stream: "lifecycle",
+      data: {
+        phase: "end",
+        startedAt: 1000,
+        endedAt: 2000,
+      },
+    });
+
+    await new Promise((resolve) => setTimeout(resolve, 0));
+    await new Promise((resolve) => setTimeout(resolve, 0));
+    await new Promise((resolve) => setTimeout(resolve, 0));
+
+    const agentCalls = calls.filter((call) => call.method === "agent");
+    expect(agentCalls).toHaveLength(2);
+    const announceParams = agentCalls[1]?.params as
+      | { accountId?: string; channel?: string; deliver?: boolean }
+      | undefined;
+    expect(announceParams?.deliver).toBe(true);
+    expect(announceParams?.channel).toBe("whatsapp");
+    expect(announceParams?.accountId).toBe("kev");
+  });
 });

src/agents/pi-embedded-utils.ts

@@ -9,7 +9,7 @@ import { formatToolDetail, resolveToolDisplay } from "./tool-display.js";
  * - <invoke name="...">...</invoke> blocks
  * - </minimax:tool_call> closing tags
  */
-function stripMinimaxToolCallXml(text: string): string {
+export function stripMinimaxToolCallXml(text: string): string {
   if (!text) return text;
   if (!/minimax:tool_call/i.test(text)) return text;
 
@@ -28,7 +28,7 @@ function stripMinimaxToolCallXml(text: string): string {
  * downgraded to text blocks like `[Tool Call: name (ID: ...)]`. These should
  * not be shown to users.
  */
-function stripDowngradedToolCallText(text: string): string {
+export function stripDowngradedToolCallText(text: string): string {
   if (!text) return text;
   if (!/\[Tool (?:Call|Result)/i.test(text)) return text;
 
@@ -165,7 +165,7 @@ function stripDowngradedToolCallText(text: string): string {
  * This is a safety net for cases where the model outputs <think> tags
  * that slip through other filtering mechanisms.
  */
-function stripThinkingTagsFromText(text: string): string {
+export function stripThinkingTagsFromText(text: string): string {
   if (!text) return text;
   // Quick check to avoid regex overhead when no tags present.
   if (!/(?:think(?:ing)?|thought|antthinking)/i.test(text)) return text;

src/agents/pi-extensions/compaction-safeguard.ts

@@ -15,15 +15,6 @@ const TURN_PREFIX_INSTRUCTIONS =
 const MAX_TOOL_FAILURES = 8;
 const MAX_TOOL_FAILURE_CHARS = 240;
 
-function isAbortError(err: unknown): boolean {
-  if (!err || typeof err !== "object") return false;
-  const name = "name" in err ? String(err.name) : "";
-  if (name === "AbortError") return true;
-  const message =
-    "message" in err && typeof err.message === "string" ? err.message.toLowerCase() : "";
-  return message.includes("aborted");
-}
-
 type ToolFailure = {
   toolCallId: string;
   toolName: string;
@@ -260,9 +251,6 @@ async function summarizeWithFallback(params: {
   try {
     return await summarizeChunks(params);
   } catch (fullError) {
-    if (params.signal.aborted || isAbortError(fullError)) {
-      throw fullError;
-    }
     console.warn(
       `Full summarization failed, trying partial: ${
         fullError instanceof Error ? fullError.message : String(fullError)
@@ -295,9 +283,6 @@ async function summarizeWithFallback(params: {
       const notes = oversizedNotes.length > 0 ? `\n\n${oversizedNotes.join("\n")}` : "";
       return partialSummary + notes;
     } catch (partialError) {
-      if (params.signal.aborted || isAbortError(partialError)) {
-        throw partialError;
-      }
       console.warn(
         `Partial summarization also failed: ${
           partialError instanceof Error ? partialError.message : String(partialError)

src/agents/tools/sessions-helpers.test.ts

@@ -0,0 +1,34 @@
+import { describe, expect, it } from "vitest";
+
+import { extractAssistantText, sanitizeTextContent } from "./sessions-helpers.js";
+
+describe("sanitizeTextContent", () => {
+  it("strips minimax tool call XML and downgraded markers", () => {
+    const input =
+      'Hello <invoke name="tool">payload</invoke></minimax:tool_call> ' +
+      "[Tool Call: foo (ID: 1)] world";
+    const result = sanitizeTextContent(input).trim();
+    expect(result).toBe("Hello  world");
+    expect(result).not.toContain("invoke");
+    expect(result).not.toContain("Tool Call");
+  });
+
+  it("strips thinking tags", () => {
+    const input = "Before <think>secret</think> after";
+    const result = sanitizeTextContent(input).trim();
+    expect(result).toBe("Before  after");
+  });
+});
+
+describe("extractAssistantText", () => {
+  it("sanitizes blocks without injecting newlines", () => {
+    const message = {
+      role: "assistant",
+      content: [
+        { type: "text", text: "Hi " },
+        { type: "text", text: "<think>secret</think>there" },
+      ],
+    };
+    expect(extractAssistantText(message)).toBe("Hi there");
+  });
+});

src/agents/tools/sessions-helpers.ts

@@ -1,4 +1,10 @@
 import type { ClawdbotConfig } from "../../config/config.js";
+import { sanitizeUserFacingText } from "../pi-embedded-helpers.js";
+import {
+  stripDowngradedToolCallText,
+  stripMinimaxToolCallXml,
+  stripThinkingTagsFromText,
+} from "../pi-embedded-utils.js";
 import { normalizeMainKey } from "../../routing/session-key.js";
 
 export type SessionKind = "main" | "group" | "cron" | "hook" | "node" | "other";
@@ -100,6 +106,15 @@ export function stripToolMessages(messages: unknown[]): unknown[] {
   });
 }
 
+/**
+ * Sanitize text content to strip tool call markers and thinking tags.
+ * This ensures user-facing text doesn't leak internal tool representations.
+ */
+export function sanitizeTextContent(text: string): string {
+  if (!text) return text;
+  return stripThinkingTagsFromText(stripDowngradedToolCallText(stripMinimaxToolCallXml(text)));
+}
+
 export function extractAssistantText(message: unknown): string | undefined {
   if (!message || typeof message !== "object") return undefined;
   if ((message as { role?: unknown }).role !== "assistant") return undefined;
@@ -110,10 +125,13 @@ export function extractAssistantText(message: unknown): string | undefined {
     if (!block || typeof block !== "object") continue;
     if ((block as { type?: unknown }).type !== "text") continue;
     const text = (block as { text?: unknown }).text;
-    if (typeof text === "string" && text.trim()) {
-      chunks.push(text);
+    if (typeof text === "string") {
+      const sanitized = sanitizeTextContent(text);
+      if (sanitized.trim()) {
+        chunks.push(sanitized);
+      }
     }
   }
   const joined = chunks.join("").trim();
-  return joined ? joined : undefined;
+  return joined ? sanitizeUserFacingText(joined) : undefined;
 }

src/auto-reply/reply/commands-subagents.test.ts

@@ -0,0 +1,23 @@
+import { describe, expect, it } from "vitest";
+
+import { extractMessageText } from "./commands-subagents.js";
+
+describe("extractMessageText", () => {
+  it("preserves user text that looks like tool call markers", () => {
+    const message = {
+      role: "user",
+      content: "Here [Tool Call: foo (ID: 1)] ok",
+    };
+    const result = extractMessageText(message);
+    expect(result?.text).toContain("[Tool Call: foo (ID: 1)]");
+  });
+
+  it("sanitizes assistant tool call markers", () => {
+    const message = {
+      role: "assistant",
+      content: "Here [Tool Call: foo (ID: 1)] ok",
+    };
+    const result = extractMessageText(message);
+    expect(result?.text).toBe("Here ok");
+  });
+});

src/auto-reply/reply/commands-subagents.ts

@@ -7,6 +7,7 @@ import {
   extractAssistantText,
   resolveInternalSessionKey,
   resolveMainSessionAlias,
+  sanitizeTextContent,
   stripToolMessages,
 } from "../../agents/tools/sessions-helpers.js";
 import type { SubagentRunRecord } from "../../agents/subagent-registry.js";
@@ -106,11 +107,14 @@ function normalizeMessageText(text: string) {
   return text.replace(/\s+/g, " ").trim();
 }
 
-function extractMessageText(message: ChatMessage): { role: string; text: string } | null {
+export function extractMessageText(message: ChatMessage): { role: string; text: string } | null {
   const role = typeof message.role === "string" ? message.role : "";
+  const shouldSanitize = role === "assistant";
   const content = message.content;
   if (typeof content === "string") {
-    const normalized = normalizeMessageText(content);
+    const normalized = normalizeMessageText(
+      shouldSanitize ? sanitizeTextContent(content) : content,
+    );
     return normalized ? { role, text: normalized } : null;
   }
   if (!Array.isArray(content)) return null;
@@ -119,8 +123,11 @@ function extractMessageText(message: ChatMessage): { role: string; text: string
     if (!block || typeof block !== "object") continue;
     if ((block as { type?: unknown }).type !== "text") continue;
     const text = (block as { text?: unknown }).text;
-    if (typeof text === "string" && text.trim()) {
-      chunks.push(text);
+    if (typeof text === "string") {
+      const value = shouldSanitize ? sanitizeTextContent(text) : text;
+      if (value.trim()) {
+        chunks.push(value);
+      }
     }
   }
   const joined = normalizeMessageText(chunks.join(" "));

src/channels/plugins/onboarding/discord.ts

@@ -47,7 +47,7 @@ async function noteDiscordTokenHelp(prompter: WizardPrompter): Promise<void> {
       "1) Discord Developer Portal → Applications → New Application",
       "2) Bot → Add Bot → Reset Token → copy token",
       "3) OAuth2 → URL Generator → scope 'bot' → invite to your server",
-      "Tip: enable Message Content Intent if you need message text.",
+      "Tip: enable Message Content Intent if you need message text. (Bot → Privileged Gateway Intents → Message Content Intent)",
       `Docs: ${formatDocsLink("/discord", "discord")}`,
     ].join("\n"),
     "Discord bot token",

src/cli/program/register.onboard.ts

@@ -56,9 +56,12 @@ export function registerOnboardCommand(program: Command) {
     )
     .option(
       "--token-provider <id>",
-      "Token provider id (non-interactive; used with --auth-choice token)",
+      "Token provider id (non-interactive; used with --auth-choice token or apiKey)",
+    )
+    .option(
+      "--token <token>",
+      "Token value (non-interactive; used with --auth-choice token or apiKey)",
     )
-    .option("--token <token>", "Token value (non-interactive; used with --auth-choice token)")
     .option(
       "--token-profile-id <id>",
       "Auth profile id (non-interactive; default: <provider>:manual)",

src/commands/auth-choice-options.ts

@@ -46,7 +46,7 @@ const AUTH_CHOICE_GROUP_DEFS: {
     value: "anthropic",
     label: "Anthropic",
     hint: "Claude Code CLI + API key",
-    choices: ["claude-cli", "token", "apiKey"],
+    choices: ["token", "claude-cli", "apiKey"],
   },
   {
     value: "minimax",

src/commands/auth-choice.api-key-flags.test.ts

@@ -0,0 +1,183 @@
+import fs from "node:fs/promises";
+import os from "node:os";
+import path from "node:path";
+
+import { afterEach, describe, expect, it, vi } from "vitest";
+
+import type { RuntimeEnv } from "../runtime.js";
+import type { WizardPrompter } from "../wizard/prompts.js";
+import { AUTH_STORE_VERSION } from "../agents/auth-profiles/constants.js";
+import { applyAuthChoice } from "./auth-choice.js";
+
+const noopAsync = async () => {};
+const noop = () => {};
+const authProfilePathFor = (agentDir: string) => path.join(agentDir, "auth-profiles.json");
+
+const previousStateDir = process.env.CLAWDBOT_STATE_DIR;
+const previousAgentDir = process.env.CLAWDBOT_AGENT_DIR;
+const previousPiAgentDir = process.env.PI_CODING_AGENT_DIR;
+const previousAnthropicKey = process.env.ANTHROPIC_API_KEY;
+const previousOpenaiKey = process.env.OPENAI_API_KEY;
+let tempStateDir: string | null = null;
+
+async function setupTempState() {
+  tempStateDir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-auth-"));
+  process.env.CLAWDBOT_STATE_DIR = tempStateDir;
+  process.env.CLAWDBOT_AGENT_DIR = path.join(tempStateDir, "agent");
+  process.env.PI_CODING_AGENT_DIR = process.env.CLAWDBOT_AGENT_DIR;
+  await fs.mkdir(process.env.CLAWDBOT_AGENT_DIR, { recursive: true });
+}
+
+function buildPrompter() {
+  const text = vi.fn(async () => "");
+  const confirm = vi.fn(async () => false);
+  const prompter: WizardPrompter = {
+    intro: vi.fn(noopAsync),
+    outro: vi.fn(noopAsync),
+    note: vi.fn(noopAsync),
+    select: vi.fn(async () => "" as never),
+    multiselect: vi.fn(async () => []),
+    text,
+    confirm,
+    progress: vi.fn(() => ({ update: noop, stop: noop })),
+  };
+  return { prompter, text, confirm };
+}
+
+const runtime: RuntimeEnv = {
+  log: vi.fn(),
+  error: vi.fn(),
+  exit: vi.fn((code: number) => {
+    throw new Error(`exit:${code}`);
+  }),
+};
+
+afterEach(async () => {
+  if (tempStateDir) {
+    await fs.rm(tempStateDir, { recursive: true, force: true });
+    tempStateDir = null;
+  }
+  if (previousStateDir === undefined) {
+    delete process.env.CLAWDBOT_STATE_DIR;
+  } else {
+    process.env.CLAWDBOT_STATE_DIR = previousStateDir;
+  }
+  if (previousAgentDir === undefined) {
+    delete process.env.CLAWDBOT_AGENT_DIR;
+  } else {
+    process.env.CLAWDBOT_AGENT_DIR = previousAgentDir;
+  }
+  if (previousPiAgentDir === undefined) {
+    delete process.env.PI_CODING_AGENT_DIR;
+  } else {
+    process.env.PI_CODING_AGENT_DIR = previousPiAgentDir;
+  }
+  if (previousAnthropicKey === undefined) {
+    delete process.env.ANTHROPIC_API_KEY;
+  } else {
+    process.env.ANTHROPIC_API_KEY = previousAnthropicKey;
+  }
+  if (previousOpenaiKey === undefined) {
+    delete process.env.OPENAI_API_KEY;
+  } else {
+    process.env.OPENAI_API_KEY = previousOpenaiKey;
+  }
+});
+
+describe("applyAuthChoice with apiKey flags", () => {
+  it("uses provided openrouter token when authChoice=apiKey", async () => {
+    await setupTempState();
+    const agentDir = process.env.CLAWDBOT_AGENT_DIR ?? "";
+    const authProfilePath = authProfilePathFor(agentDir);
+    await fs.writeFile(
+      authProfilePath,
+      JSON.stringify({
+        version: AUTH_STORE_VERSION,
+        profiles: {
+          "openrouter:legacy": {
+            type: "oauth",
+            provider: "openrouter",
+            access: "access",
+            refresh: "refresh",
+            expires: Date.now() + 60_000,
+          },
+        },
+      }),
+      "utf8",
+    );
+
+    const { prompter, text, confirm } = buildPrompter();
+    const result = await applyAuthChoice({
+      authChoice: "apiKey",
+      config: {},
+      prompter,
+      runtime,
+      setDefaultModel: true,
+      opts: {
+        tokenProvider: "openrouter",
+        token: "sk-openrouter-flag",
+      },
+    });
+
+    expect(text).not.toHaveBeenCalled();
+    expect(confirm).not.toHaveBeenCalled();
+    expect(result.config.auth?.profiles?.["openrouter:default"]).toMatchObject({
+      provider: "openrouter",
+      mode: "api_key",
+    });
+
+    const raw = await fs.readFile(authProfilePath, "utf8");
+    const parsed = JSON.parse(raw) as { profiles?: Record<string, { key?: string }> };
+    expect(parsed.profiles?.["openrouter:default"]?.key).toBe("sk-openrouter-flag");
+  });
+
+  it("uses provided openai token when authChoice=apiKey", async () => {
+    await setupTempState();
+    const { prompter, text, confirm } = buildPrompter();
+
+    await applyAuthChoice({
+      authChoice: "apiKey",
+      config: {},
+      prompter,
+      runtime,
+      setDefaultModel: true,
+      opts: {
+        tokenProvider: "openai",
+        token: "sk-openai-flag",
+      },
+    });
+
+    expect(text).not.toHaveBeenCalled();
+    expect(confirm).not.toHaveBeenCalled();
+    expect(process.env.OPENAI_API_KEY).toBe("sk-openai-flag");
+    const envPath = path.join(process.env.CLAWDBOT_STATE_DIR ?? "", ".env");
+    const envContents = await fs.readFile(envPath, "utf8");
+    expect(envContents).toContain("OPENAI_API_KEY=sk-openai-flag");
+  });
+
+  it("uses provided anthropic token when authChoice=apiKey", async () => {
+    await setupTempState();
+    process.env.ANTHROPIC_API_KEY = "sk-env-test";
+    const { prompter, text, confirm } = buildPrompter();
+
+    await applyAuthChoice({
+      authChoice: "apiKey",
+      config: {},
+      prompter,
+      runtime,
+      setDefaultModel: true,
+      opts: {
+        tokenProvider: "anthropic",
+        token: "sk-anthropic-flag",
+      },
+    });
+
+    expect(text).not.toHaveBeenCalled();
+    expect(confirm).not.toHaveBeenCalled();
+
+    const authProfilePath = authProfilePathFor(process.env.CLAWDBOT_AGENT_DIR ?? "");
+    const raw = await fs.readFile(authProfilePath, "utf8");
+    const parsed = JSON.parse(raw) as { profiles?: Record<string, { key?: string }> };
+    expect(parsed.profiles?.["anthropic:default"]?.key).toBe("sk-anthropic-flag");
+  });
+});

src/commands/auth-choice.apply.anthropic.ts

@@ -8,6 +8,7 @@ import {
   normalizeApiKeyInput,
   validateApiKeyInput,
 } from "./auth-choice.api-key.js";
+import { normalizeProviderId } from "../agents/model-selection.js";
 import type { ApplyAuthChoiceParams, ApplyAuthChoiceResult } from "./auth-choice.apply.js";
 import { buildTokenProfileId, validateAnthropicSetupToken } from "./auth-token.js";
 import { applyAuthProfileConfig, setAnthropicApiKey } from "./onboard-auth.js";
@@ -198,10 +199,23 @@ export async function applyAuthChoiceAnthropic(
   }
 
   if (params.authChoice === "apiKey") {
+    const tokenProvider = params.opts?.tokenProvider
+      ? normalizeProviderId(params.opts.tokenProvider)
+      : undefined;
+    const explicitToken =
+      tokenProvider === "anthropic" ? normalizeApiKeyInput(params.opts?.token ?? "") : "";
+    if (tokenProvider && tokenProvider !== "anthropic") {
+      return null;
+    }
+
     let nextConfig = params.config;
     let hasCredential = false;
     const envKey = process.env.ANTHROPIC_API_KEY?.trim();
-    if (envKey) {
+    if (explicitToken) {
+      await setAnthropicApiKey(explicitToken, params.agentDir);
+      hasCredential = true;
+    }
+    if (!hasCredential && envKey) {
       const useExisting = await params.prompter.confirm({
         message: `Use existing ANTHROPIC_API_KEY (env, ${formatApiKeyPreview(envKey)})?`,
         initialValue: true,

src/commands/auth-choice.apply.api-providers.ts

@@ -1,4 +1,5 @@
 import { ensureAuthProfileStore, resolveAuthProfileOrder } from "../agents/auth-profiles.js";
+import { normalizeProviderId } from "../agents/model-selection.js";
 import { resolveEnvApiKey } from "../agents/model-auth.js";
 import {
   formatApiKeyPreview,
@@ -6,6 +7,7 @@ import {
   validateApiKeyInput,
 } from "./auth-choice.api-key.js";
 import type { ApplyAuthChoiceParams, ApplyAuthChoiceResult } from "./auth-choice.apply.js";
+import type { AuthChoice } from "./onboard-types.js";
 import { applyDefaultModelChoice } from "./auth-choice.default-model.js";
 import {
   applyGoogleGeminiModelDefault,
@@ -56,53 +58,86 @@ export async function applyAuthChoiceApiProviders(
     );
   };
 
-  if (params.authChoice === "openrouter-api-key") {
-    const store = ensureAuthProfileStore(params.agentDir, {
-      allowKeychainPrompt: false,
-    });
-    const profileOrder = resolveAuthProfileOrder({
-      cfg: nextConfig,
-      store,
-      provider: "openrouter",
-    });
-    const existingProfileId = profileOrder.find((profileId) => Boolean(store.profiles[profileId]));
-    const existingCred = existingProfileId ? store.profiles[existingProfileId] : undefined;
+  const tokenProvider = params.opts?.tokenProvider
+    ? normalizeProviderId(params.opts.tokenProvider)
+    : undefined;
+  const tokenValue = params.opts?.token;
+  let authChoice = params.authChoice;
+  if (
+    authChoice === "apiKey" &&
+    tokenProvider &&
+    tokenProvider !== "anthropic" &&
+    tokenProvider !== "openai"
+  ) {
+    const mapped: Partial<Record<string, AuthChoice>> = {
+      openrouter: "openrouter-api-key",
+      "vercel-ai-gateway": "ai-gateway-api-key",
+      moonshot: "moonshot-api-key",
+      "kimi-code": "kimi-code-api-key",
+      google: "gemini-api-key",
+      zai: "zai-api-key",
+      synthetic: "synthetic-api-key",
+      opencode: "opencode-zen",
+    };
+    authChoice = mapped[tokenProvider] ?? authChoice;
+  }
+
+  if (authChoice === "openrouter-api-key") {
     let profileId = "openrouter:default";
     let mode: "api_key" | "oauth" | "token" = "api_key";
     let hasCredential = false;
+    const explicitToken =
+      tokenProvider === "openrouter" ? normalizeApiKeyInput(tokenValue ?? "") : "";
 
-    if (existingProfileId && existingCred?.type) {
-      profileId = existingProfileId;
-      mode =
-        existingCred.type === "oauth"
-          ? "oauth"
-          : existingCred.type === "token"
-            ? "token"
-            : "api_key";
+    if (explicitToken) {
+      await setOpenrouterApiKey(explicitToken, params.agentDir);
       hasCredential = true;
-    }
+    } else {
+      const store = ensureAuthProfileStore(params.agentDir, {
+        allowKeychainPrompt: false,
+      });
+      const profileOrder = resolveAuthProfileOrder({
+        cfg: nextConfig,
+        store,
+        provider: "openrouter",
+      });
+      const existingProfileId = profileOrder.find((profileId) =>
+        Boolean(store.profiles[profileId]),
+      );
+      const existingCred = existingProfileId ? store.profiles[existingProfileId] : undefined;
+      if (existingProfileId && existingCred?.type) {
+        profileId = existingProfileId;
+        mode =
+          existingCred.type === "oauth"
+            ? "oauth"
+            : existingCred.type === "token"
+              ? "token"
+              : "api_key";
+        hasCredential = true;
+      }
 
-    if (!hasCredential) {
-      const envKey = resolveEnvApiKey("openrouter");
-      if (envKey) {
-        const useExisting = await params.prompter.confirm({
-          message: `Use existing OPENROUTER_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
-          initialValue: true,
-        });
-        if (useExisting) {
-          await setOpenrouterApiKey(envKey.apiKey, params.agentDir);
-          hasCredential = true;
+      if (!hasCredential) {
+        const envKey = resolveEnvApiKey("openrouter");
+        if (envKey) {
+          const useExisting = await params.prompter.confirm({
+            message: `Use existing OPENROUTER_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
+            initialValue: true,
+          });
+          if (useExisting) {
+            await setOpenrouterApiKey(envKey.apiKey, params.agentDir);
+            hasCredential = true;
+          }
         }
       }
-    }
 
-    if (!hasCredential) {
-      const key = await params.prompter.text({
-        message: "Enter OpenRouter API key",
-        validate: validateApiKeyInput,
-      });
-      await setOpenrouterApiKey(normalizeApiKeyInput(String(key)), params.agentDir);
-      hasCredential = true;
+      if (!hasCredential) {
+        const key = await params.prompter.text({
+          message: "Enter OpenRouter API key",
+          validate: validateApiKeyInput,
+        });
+        await setOpenrouterApiKey(normalizeApiKeyInput(String(key)), params.agentDir);
+        hasCredential = true;
+      }
     }
 
     if (hasCredential) {
@@ -129,10 +164,16 @@ export async function applyAuthChoiceApiProviders(
     return { config: nextConfig, agentModelOverride };
   }
 
-  if (params.authChoice === "ai-gateway-api-key") {
+  if (authChoice === "ai-gateway-api-key") {
     let hasCredential = false;
+    const explicitToken =
+      tokenProvider === "vercel-ai-gateway" ? normalizeApiKeyInput(tokenValue ?? "") : "";
+    if (explicitToken) {
+      await setVercelAiGatewayApiKey(explicitToken, params.agentDir);
+      hasCredential = true;
+    }
     const envKey = resolveEnvApiKey("vercel-ai-gateway");
-    if (envKey) {
+    if (!hasCredential && envKey) {
       const useExisting = await params.prompter.confirm({
         message: `Use existing AI_GATEWAY_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
         initialValue: true,
@@ -171,10 +212,16 @@ export async function applyAuthChoiceApiProviders(
     return { config: nextConfig, agentModelOverride };
   }
 
-  if (params.authChoice === "moonshot-api-key") {
+  if (authChoice === "moonshot-api-key") {
     let hasCredential = false;
+    const explicitToken =
+      tokenProvider === "moonshot" ? normalizeApiKeyInput(tokenValue ?? "") : "";
+    if (explicitToken) {
+      await setMoonshotApiKey(explicitToken, params.agentDir);
+      hasCredential = true;
+    }
     const envKey = resolveEnvApiKey("moonshot");
-    if (envKey) {
+    if (!hasCredential && envKey) {
       const useExisting = await params.prompter.confirm({
         message: `Use existing MOONSHOT_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
         initialValue: true,
@@ -212,17 +259,25 @@ export async function applyAuthChoiceApiProviders(
     return { config: nextConfig, agentModelOverride };
   }
 
-  if (params.authChoice === "kimi-code-api-key") {
-    await params.prompter.note(
-      [
-        "Kimi Code uses a dedicated endpoint and API key.",
-        "Get your API key at: https://www.kimi.com/code/en",
-      ].join("\n"),
-      "Kimi Code",
-    );
+  if (authChoice === "kimi-code-api-key") {
     let hasCredential = false;
+    const explicitToken =
+      tokenProvider === "kimi-code" ? normalizeApiKeyInput(tokenValue ?? "") : "";
+    if (explicitToken) {
+      await setKimiCodeApiKey(explicitToken, params.agentDir);
+      hasCredential = true;
+    }
+    if (!hasCredential) {
+      await params.prompter.note(
+        [
+          "Kimi Code uses a dedicated endpoint and API key.",
+          "Get your API key at: https://www.kimi.com/code/en",
+        ].join("\n"),
+        "Kimi Code",
+      );
+    }
     const envKey = resolveEnvApiKey("kimi-code");
-    if (envKey) {
+    if (!hasCredential && envKey) {
       const useExisting = await params.prompter.confirm({
         message: `Use existing KIMICODE_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
         initialValue: true,
@@ -261,10 +316,15 @@ export async function applyAuthChoiceApiProviders(
     return { config: nextConfig, agentModelOverride };
   }
 
-  if (params.authChoice === "gemini-api-key") {
+  if (authChoice === "gemini-api-key") {
     let hasCredential = false;
+    const explicitToken = tokenProvider === "google" ? normalizeApiKeyInput(tokenValue ?? "") : "";
+    if (explicitToken) {
+      await setGeminiApiKey(explicitToken, params.agentDir);
+      hasCredential = true;
+    }
     const envKey = resolveEnvApiKey("google");
-    if (envKey) {
+    if (!hasCredential && envKey) {
       const useExisting = await params.prompter.confirm({
         message: `Use existing GEMINI_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
         initialValue: true,
@@ -302,10 +362,15 @@ export async function applyAuthChoiceApiProviders(
     return { config: nextConfig, agentModelOverride };
   }
 
-  if (params.authChoice === "zai-api-key") {
+  if (authChoice === "zai-api-key") {
     let hasCredential = false;
+    const explicitToken = tokenProvider === "zai" ? normalizeApiKeyInput(tokenValue ?? "") : "";
+    if (explicitToken) {
+      await setZaiApiKey(explicitToken, params.agentDir);
+      hasCredential = true;
+    }
     const envKey = resolveEnvApiKey("zai");
-    if (envKey) {
+    if (!hasCredential && envKey) {
       const useExisting = await params.prompter.confirm({
         message: `Use existing ZAI_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
         initialValue: true,
@@ -359,12 +424,18 @@ export async function applyAuthChoiceApiProviders(
     return { config: nextConfig, agentModelOverride };
   }
 
-  if (params.authChoice === "synthetic-api-key") {
-    const key = await params.prompter.text({
-      message: "Enter Synthetic API key",
-      validate: (value) => (value?.trim() ? undefined : "Required"),
-    });
-    await setSyntheticApiKey(String(key).trim(), params.agentDir);
+  if (authChoice === "synthetic-api-key") {
+    const explicitToken =
+      tokenProvider === "synthetic" ? normalizeApiKeyInput(tokenValue ?? "") : "";
+    if (explicitToken) {
+      await setSyntheticApiKey(explicitToken, params.agentDir);
+    } else {
+      const key = await params.prompter.text({
+        message: "Enter Synthetic API key",
+        validate: (value) => (value?.trim() ? undefined : "Required"),
+      });
+      await setSyntheticApiKey(String(key).trim(), params.agentDir);
+    }
     nextConfig = applyAuthProfileConfig(nextConfig, {
       profileId: "synthetic:default",
       provider: "synthetic",
@@ -387,18 +458,26 @@ export async function applyAuthChoiceApiProviders(
     return { config: nextConfig, agentModelOverride };
   }
 
-  if (params.authChoice === "opencode-zen") {
-    await params.prompter.note(
-      [
-        "OpenCode Zen provides access to Claude, GPT, Gemini, and more models.",
-        "Get your API key at: https://opencode.ai/auth",
-        "Requires an active OpenCode Zen subscription.",
-      ].join("\n"),
-      "OpenCode Zen",
-    );
+  if (authChoice === "opencode-zen") {
     let hasCredential = false;
+    const explicitToken =
+      tokenProvider === "opencode" ? normalizeApiKeyInput(tokenValue ?? "") : "";
+    if (explicitToken) {
+      await setOpencodeZenApiKey(explicitToken, params.agentDir);
+      hasCredential = true;
+    }
+    if (!hasCredential) {
+      await params.prompter.note(
+        [
+          "OpenCode Zen provides access to Claude, GPT, Gemini, and more models.",
+          "Get your API key at: https://opencode.ai/auth",
+          "Requires an active OpenCode Zen subscription.",
+        ].join("\n"),
+        "OpenCode Zen",
+      );
+    }
     const envKey = resolveEnvApiKey("opencode");
-    if (envKey) {
+    if (!hasCredential && envKey) {
       const useExisting = await params.prompter.confirm({
         message: `Use existing OPENCODE_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
         initialValue: true,

src/commands/auth-choice.apply.openai.ts

@@ -1,5 +1,6 @@
 import { loginOpenAICodex } from "@mariozechner/pi-ai";
 import { CODEX_CLI_PROFILE_ID, ensureAuthProfileStore } from "../agents/auth-profiles.js";
+import { normalizeProviderId } from "../agents/model-selection.js";
 import { resolveEnvApiKey } from "../agents/model-auth.js";
 import { upsertSharedEnvVar } from "../infra/env-file.js";
 import { isRemoteEnvironment } from "./oauth-env.js";
@@ -20,33 +21,47 @@ import {
 export async function applyAuthChoiceOpenAI(
   params: ApplyAuthChoiceParams,
 ): Promise<ApplyAuthChoiceResult | null> {
-  if (params.authChoice === "openai-api-key") {
-    const envKey = resolveEnvApiKey("openai");
-    if (envKey) {
-      const useExisting = await params.prompter.confirm({
-        message: `Use existing OPENAI_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
-        initialValue: true,
-      });
-      if (useExisting) {
-        const result = upsertSharedEnvVar({
-          key: "OPENAI_API_KEY",
-          value: envKey.apiKey,
+  const tokenProvider = params.opts?.tokenProvider
+    ? normalizeProviderId(params.opts.tokenProvider)
+    : undefined;
+  const explicitToken =
+    tokenProvider === "openai" ? normalizeApiKeyInput(params.opts?.token ?? "") : "";
+  let authChoice = params.authChoice;
+  if (authChoice === "apiKey" && tokenProvider === "openai") {
+    authChoice = "openai-api-key";
+  }
+
+  if (authChoice === "openai-api-key") {
+    if (!explicitToken) {
+      const envKey = resolveEnvApiKey("openai");
+      if (envKey) {
+        const useExisting = await params.prompter.confirm({
+          message: `Use existing OPENAI_API_KEY (${envKey.source}, ${formatApiKeyPreview(envKey.apiKey)})?`,
+          initialValue: true,
         });
-        if (!process.env.OPENAI_API_KEY) {
-          process.env.OPENAI_API_KEY = envKey.apiKey;
+        if (useExisting) {
+          const result = upsertSharedEnvVar({
+            key: "OPENAI_API_KEY",
+            value: envKey.apiKey,
+          });
+          if (!process.env.OPENAI_API_KEY) {
+            process.env.OPENAI_API_KEY = envKey.apiKey;
+          }
+          await params.prompter.note(
+            `Copied OPENAI_API_KEY to ${result.path} for launchd compatibility.`,
+            "OpenAI API key",
+          );
+          return { config: params.config };
         }
-        await params.prompter.note(
-          `Copied OPENAI_API_KEY to ${result.path} for launchd compatibility.`,
-          "OpenAI API key",
-        );
-        return { config: params.config };
       }
     }
 
-    const key = await params.prompter.text({
-      message: "Enter OpenAI API key",
-      validate: validateApiKeyInput,
-    });
+    const key = explicitToken
+      ? explicitToken
+      : await params.prompter.text({
+          message: "Enter OpenAI API key",
+          validate: validateApiKeyInput,
+        });
     const trimmed = normalizeApiKeyInput(String(key));
     const result = upsertSharedEnvVar({
       key: "OPENAI_API_KEY",

src/commands/auth-choice.apply.ts

@@ -21,6 +21,10 @@ export type ApplyAuthChoiceParams = {
   agentDir?: string;
   setDefaultModel: boolean;
   agentId?: string;
+  opts?: {
+    tokenProvider?: string;
+    token?: string;
+  };
 };
 
 export type ApplyAuthChoiceResult = {

src/commands/onboard-helpers.test.ts

@@ -1,4 +1,4 @@
-import { describe, expect, it, vi } from "vitest";
+import { afterEach, describe, expect, it, vi } from "vitest";
 
 import { openUrl, resolveBrowserOpenCommand, resolveControlUiLinks } from "./onboard-helpers.js";
 
@@ -21,9 +21,17 @@ vi.mock("../infra/tailnet.js", () => ({
   pickPrimaryTailnetIPv4: mocks.pickPrimaryTailnetIPv4,
 }));
 
+afterEach(() => {
+  vi.unstubAllEnvs();
+});
+
 describe("openUrl", () => {
   it("quotes URLs on win32 so '&' is not treated as cmd separator", async () => {
-    vi.spyOn(process, "platform", "get").mockReturnValue("win32");
+    vi.stubEnv("VITEST", "");
+    vi.stubEnv("NODE_ENV", "");
+    const platformSpy = vi.spyOn(process, "platform", "get").mockReturnValue("win32");
+    vi.stubEnv("VITEST", "");
+    vi.stubEnv("NODE_ENV", "development");
 
     const url =
       "https://accounts.google.com/o/oauth2/v2/auth?client_id=abc&response_type=code&redirect_uri=http%3A%2F%2Flocalhost";
@@ -39,15 +47,18 @@ describe("openUrl", () => {
       timeoutMs: 5_000,
       windowsVerbatimArguments: true,
     });
+
+    platformSpy.mockRestore();
   });
 });
 
 describe("resolveBrowserOpenCommand", () => {
   it("marks win32 commands as quoteUrl=true", async () => {
-    vi.spyOn(process, "platform", "get").mockReturnValue("win32");
+    const platformSpy = vi.spyOn(process, "platform", "get").mockReturnValue("win32");
     const resolved = await resolveBrowserOpenCommand();
     expect(resolved.argv).toEqual(["cmd", "/c", "start", ""]);
     expect(resolved.quoteUrl).toBe(true);
+    platformSpy.mockRestore();
   });
 });
 

src/commands/onboard-helpers.ts

@@ -192,6 +192,7 @@ function resolveSshTargetHint(): string {
 }
 
 export async function openUrl(url: string): Promise<boolean> {
+  if (shouldSkipBrowserOpenInTests()) return false;
   const resolved = await resolveBrowserOpenCommand();
   if (!resolved.argv) return false;
   const quoteUrl = resolved.quoteUrl === true;
@@ -218,6 +219,7 @@ export async function openUrl(url: string): Promise<boolean> {
 }
 
 export async function openUrlInBackground(url: string): Promise<boolean> {
+  if (shouldSkipBrowserOpenInTests()) return false;
   if (process.platform !== "darwin") return false;
   const resolved = await resolveBrowserOpenCommand();
   if (!resolved.argv || resolved.command !== "open") return false;
@@ -308,6 +310,11 @@ export async function detectBinary(name: string): Promise<boolean> {
   }
 }
 
+function shouldSkipBrowserOpenInTests(): boolean {
+  if (process.env.VITEST) return true;
+  return process.env.NODE_ENV === "test";
+}
+
 export async function probeGatewayReachable(params: {
   url: string;
   token?: string;

src/commands/onboard-non-interactive.gateway.test.ts

@@ -4,18 +4,43 @@ import os from "node:os";
 import path from "node:path";
 
 import { afterAll, beforeAll, describe, expect, it, vi } from "vitest";
-import { WebSocket } from "ws";
 
-import {
-  loadOrCreateDeviceIdentity,
-  publicKeyRawBase64UrlFromPem,
-  signDevicePayload,
-} from "../infra/device-identity.js";
-import { buildDeviceAuthPayload } from "../gateway/device-auth.js";
-import { PROTOCOL_VERSION } from "../gateway/protocol/index.js";
-import { rawDataToString } from "../infra/ws.js";
 import { getDeterministicFreePortBlock } from "../test-utils/ports.js";
-import { GATEWAY_CLIENT_MODES, GATEWAY_CLIENT_NAMES } from "../utils/message-channel.js";
+
+const gatewayClientCalls: Array<{
+  url?: string;
+  token?: string;
+  password?: string;
+  onHelloOk?: () => void;
+  onClose?: (code: number, reason: string) => void;
+}> = [];
+
+vi.mock("../gateway/client.js", () => ({
+  GatewayClient: class {
+    params: {
+      url?: string;
+      token?: string;
+      password?: string;
+      onHelloOk?: () => void;
+    };
+    constructor(params: {
+      url?: string;
+      token?: string;
+      password?: string;
+      onHelloOk?: () => void;
+    }) {
+      this.params = params;
+      gatewayClientCalls.push(params);
+    }
+    async request() {
+      return { ok: true };
+    }
+    start() {
+      queueMicrotask(() => this.params.onHelloOk?.());
+    }
+    stop() {}
+  },
+}));
 
 async function getFreePort(): Promise<number> {
   return await new Promise((resolve, reject) => {
@@ -41,85 +66,6 @@ async function getFreeGatewayPort(): Promise<number> {
   return await getDeterministicFreePortBlock({ offsets: [0, 1, 2, 4] });
 }
 
-async function onceMessage<T = unknown>(
-  ws: WebSocket,
-  filter: (obj: unknown) => boolean,
-  timeoutMs = 5000,
-): Promise<T> {
-  return await new Promise<T>((resolve, reject) => {
-    const timer = setTimeout(() => reject(new Error("timeout")), timeoutMs);
-    const closeHandler = (code: number, reason: Buffer) => {
-      clearTimeout(timer);
-      ws.off("message", handler);
-      reject(new Error(`closed ${code}: ${rawDataToString(reason)}`));
-    };
-    const handler = (data: WebSocket.RawData) => {
-      const obj = JSON.parse(rawDataToString(data));
-      if (!filter(obj)) return;
-      clearTimeout(timer);
-      ws.off("message", handler);
-      ws.off("close", closeHandler);
-      resolve(obj as T);
-    };
-    ws.on("message", handler);
-    ws.once("close", closeHandler);
-  });
-}
-
-async function connectReq(params: { url: string; token?: string }) {
-  const ws = new WebSocket(params.url);
-  await new Promise<void>((resolve) => ws.once("open", resolve));
-  const identity = loadOrCreateDeviceIdentity();
-  const signedAtMs = Date.now();
-  const payload = buildDeviceAuthPayload({
-    deviceId: identity.deviceId,
-    clientId: GATEWAY_CLIENT_NAMES.TEST,
-    clientMode: GATEWAY_CLIENT_MODES.TEST,
-    role: "operator",
-    scopes: [],
-    signedAtMs,
-    token: params.token ?? null,
-  });
-  const device = {
-    id: identity.deviceId,
-    publicKey: publicKeyRawBase64UrlFromPem(identity.publicKeyPem),
-    signature: signDevicePayload(identity.privateKeyPem, payload),
-    signedAt: signedAtMs,
-  };
-  ws.send(
-    JSON.stringify({
-      type: "req",
-      id: "c1",
-      method: "connect",
-      params: {
-        minProtocol: PROTOCOL_VERSION,
-        maxProtocol: PROTOCOL_VERSION,
-        client: {
-          id: GATEWAY_CLIENT_NAMES.TEST,
-          displayName: "vitest",
-          version: "dev",
-          platform: process.platform,
-          mode: GATEWAY_CLIENT_MODES.TEST,
-        },
-        caps: [],
-        auth: params.token ? { token: params.token } : undefined,
-        device,
-      },
-    }),
-  );
-  const res = await onceMessage<{
-    type: "res";
-    id: string;
-    ok: boolean;
-    error?: { message?: string };
-  }>(ws, (o) => {
-    const obj = o as { type?: unknown; id?: unknown } | undefined;
-    return obj?.type === "res" && obj?.id === "c1";
-  });
-  ws.close();
-  return res;
-}
-
 const runtime = {
   log: () => {},
   error: (msg: string) => {
@@ -152,7 +98,6 @@ describe("onboard (non-interactive): gateway and remote auth", () => {
     const stateDir = await fs.mkdtemp(path.join(tempHome, prefix));
     process.env.CLAWDBOT_STATE_DIR = stateDir;
     delete process.env.CLAWDBOT_CONFIG_PATH;
-    vi.resetModules();
     return stateDir;
   };
 
@@ -207,8 +152,9 @@ describe("onboard (non-interactive): gateway and remote auth", () => {
       runtime,
     );
 
-    const { CONFIG_PATH_CLAWDBOT } = await import("../config/config.js");
-    const cfg = JSON.parse(await fs.readFile(CONFIG_PATH_CLAWDBOT, "utf8")) as {
+    const { resolveConfigPath } = await import("../config/paths.js");
+    const configPath = resolveConfigPath(process.env, stateDir);
+    const cfg = JSON.parse(await fs.readFile(configPath, "utf8")) as {
       gateway?: { auth?: { mode?: string; token?: string } };
       agents?: { defaults?: { workspace?: string } };
     };
@@ -217,25 +163,12 @@ describe("onboard (non-interactive): gateway and remote auth", () => {
     expect(cfg?.gateway?.auth?.mode).toBe("token");
     expect(cfg?.gateway?.auth?.token).toBe(token);
 
-    const { startGatewayServer } = await import("../gateway/server.js");
-    const port = await getFreePort();
-    const server = await startGatewayServer(port, {
-      bind: "loopback",
-      controlUiEnabled: false,
-    });
-    try {
-      const resNoToken = await connectReq({ url: `ws://127.0.0.1:${port}` });
-      expect(resNoToken.ok).toBe(false);
-      expect(resNoToken.error?.message ?? "").toContain("unauthorized");
-
-      const resToken = await connectReq({
-        url: `ws://127.0.0.1:${port}`,
-        token,
-      });
-      expect(resToken.ok).toBe(true);
-    } finally {
-      await server.close({ reason: "non-interactive onboard auth test" });
-    }
+    const { authorizeGatewayConnect, resolveGatewayAuth } = await import("../gateway/auth.js");
+    const auth = resolveGatewayAuth({ authConfig: cfg.gateway?.auth, env: process.env });
+    const resNoToken = await authorizeGatewayConnect({ auth, connectAuth: { token: undefined } });
+    expect(resNoToken.ok).toBe(false);
+    const resToken = await authorizeGatewayConnect({ auth, connectAuth: { token } });
+    expect(resToken.ok).toBe(true);
 
     await fs.rm(stateDir, { recursive: true, force: true });
   }, 60_000);
@@ -244,43 +177,37 @@ describe("onboard (non-interactive): gateway and remote auth", () => {
     const stateDir = await initStateDir("state-remote-");
     const port = await getFreePort();
     const token = "tok_remote_123";
-    const { startGatewayServer } = await import("../gateway/server.js");
-    const server = await startGatewayServer(port, {
-      bind: "loopback",
-      auth: { mode: "token", token },
-      controlUiEnabled: false,
-    });
+    const { runNonInteractiveOnboarding } = await import("./onboard-non-interactive.js");
+    await runNonInteractiveOnboarding(
+      {
+        nonInteractive: true,
+        mode: "remote",
+        remoteUrl: `ws://127.0.0.1:${port}`,
+        remoteToken: token,
+        authChoice: "skip",
+        json: true,
+      },
+      runtime,
+    );
 
-    try {
-      const { runNonInteractiveOnboarding } = await import("./onboard-non-interactive.js");
-      await runNonInteractiveOnboarding(
-        {
-          nonInteractive: true,
-          mode: "remote",
-          remoteUrl: `ws://127.0.0.1:${port}`,
-          remoteToken: token,
-          authChoice: "skip",
-          json: true,
-        },
-        runtime,
-      );
+    const { resolveConfigPath } = await import("../config/config.js");
+    const cfg = JSON.parse(await fs.readFile(resolveConfigPath(), "utf8")) as {
+      gateway?: { mode?: string; remote?: { url?: string; token?: string } };
+    };
 
-      const { resolveConfigPath } = await import("../config/config.js");
-      const cfg = JSON.parse(await fs.readFile(resolveConfigPath(), "utf8")) as {
-        gateway?: { mode?: string; remote?: { url?: string; token?: string } };
-      };
+    expect(cfg.gateway?.mode).toBe("remote");
+    expect(cfg.gateway?.remote?.url).toBe(`ws://127.0.0.1:${port}`);
+    expect(cfg.gateway?.remote?.token).toBe(token);
 
-      expect(cfg.gateway?.mode).toBe("remote");
-      expect(cfg.gateway?.remote?.url).toBe(`ws://127.0.0.1:${port}`);
-      expect(cfg.gateway?.remote?.token).toBe(token);
+    gatewayClientCalls.length = 0;
+    const { callGateway } = await import("../gateway/call.js");
+    const health = await callGateway<{ ok?: boolean }>({ method: "health" });
+    expect(health?.ok).toBe(true);
+    const lastCall = gatewayClientCalls[gatewayClientCalls.length - 1];
+    expect(lastCall?.url).toBe(`ws://127.0.0.1:${port}`);
+    expect(lastCall?.token).toBe(token);
 
-      const { callGateway } = await import("../gateway/call.js");
-      const health = await callGateway<{ ok?: boolean }>({ method: "health" });
-      expect(health?.ok).toBe(true);
-    } finally {
-      await server.close({ reason: "non-interactive remote test complete" });
-      await fs.rm(stateDir, { recursive: true, force: true });
-    }
+    await fs.rm(stateDir, { recursive: true, force: true });
   }, 60_000);
 
   it("auto-enables token auth when binding LAN and persists the token", async () => {
@@ -336,27 +263,12 @@ describe("onboard (non-interactive): gateway and remote auth", () => {
     const token = cfg.gateway?.auth?.token ?? "";
     expect(token.length).toBeGreaterThan(8);
 
-    const { startGatewayServer } = await import("../gateway/server.js");
-    const server = await startGatewayServer(port, {
-      controlUiEnabled: false,
-      auth: {
-        mode: "token",
-        token,
-      },
-    });
-    try {
-      const resNoToken = await connectReq({ url: `ws://127.0.0.1:${port}` });
-      expect(resNoToken.ok).toBe(false);
-      expect(resNoToken.error?.message ?? "").toContain("unauthorized");
-
-      const resToken = await connectReq({
-        url: `ws://127.0.0.1:${port}`,
-        token,
-      });
-      expect(resToken.ok).toBe(true);
-    } finally {
-      await server.close({ reason: "lan auto-token test complete" });
-    }
+    const { authorizeGatewayConnect, resolveGatewayAuth } = await import("../gateway/auth.js");
+    const auth = resolveGatewayAuth({ authConfig: cfg.gateway?.auth, env: process.env });
+    const resNoToken = await authorizeGatewayConnect({ auth, connectAuth: { token: undefined } });
+    expect(resNoToken.ok).toBe(false);
+    const resToken = await authorizeGatewayConnect({ auth, connectAuth: { token } });
+    expect(resToken.ok).toBe(true);
 
     await fs.rm(stateDir, { recursive: true, force: true });
   }, 60_000);

src/commands/onboard-types.ts

@@ -50,9 +50,9 @@ export type OnboardOptions = {
   acceptRisk?: boolean;
   reset?: boolean;
   authChoice?: AuthChoice;
-  /** Used when `authChoice=token` in non-interactive mode. */
+  /** Used when `authChoice=token` in non-interactive mode, or `authChoice=apiKey` in wizard mode. */
   tokenProvider?: string;
-  /** Used when `authChoice=token` in non-interactive mode. */
+  /** Used when `authChoice=token` in non-interactive mode, or `authChoice=apiKey` in wizard mode. */
   token?: string;
   /** Used when `authChoice=token` in non-interactive mode. */
   tokenProfileId?: string;

src/discord/monitor.tool-result.accepts-guild-messages-mentionpatterns-match.test.ts

@@ -36,6 +36,7 @@ vi.mock("../config/sessions.js", async (importOriginal) => {
 });
 
 beforeEach(() => {
+  vi.useRealTimers();
   sendMock.mockReset().mockResolvedValue(undefined);
   updateLastRouteMock.mockReset();
   dispatchMock.mockReset().mockImplementation(async ({ dispatcher }) => {

src/gateway/openai-http.e2e.test.ts

@@ -1,11 +1,23 @@
-import { describe, expect, it } from "vitest";
+import { afterAll, beforeAll, describe, expect, it } from "vitest";
 
 import { HISTORY_CONTEXT_MARKER } from "../auto-reply/reply/history.js";
 import { CURRENT_MESSAGE_MARKER } from "../auto-reply/reply/mentions.js";
 import { emitAgentEvent } from "../infra/agent-events.js";
 import { agentCommand, getFreePort, installGatewayTestHooks } from "./test-helpers.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
+
+let enabledServer: Awaited<ReturnType<typeof startServer>>;
+let enabledPort: number;
+
+beforeAll(async () => {
+  enabledPort = await getFreePort();
+  enabledServer = await startServer(enabledPort);
+});
+
+afterAll(async () => {
+  await enabledServer.close({ reason: "openai http enabled suite done" });
+});
 
 async function startServerWithDefaultConfig(port: number) {
   const { startGatewayServer } = await import("./server.js");
@@ -82,8 +94,7 @@ describe("OpenAI-compatible HTTP API (e2e)", () => {
   });
 
   it("handles request validation and routing", async () => {
-    const port = await getFreePort();
-    const server = await startServer(port);
+    const port = enabledPort;
     const mockAgentOnce = (payloads: Array<{ text: string }>) => {
       agentCommand.mockReset();
       agentCommand.mockResolvedValueOnce({ payloads } as never);
@@ -330,13 +341,12 @@ describe("OpenAI-compatible HTTP API (e2e)", () => {
         );
       }
     } finally {
-      await server.close({ reason: "test done" });
+      // shared server
     }
   });
 
   it("streams SSE chunks when stream=true", async () => {
-    const port = await getFreePort();
-    const server = await startServer(port);
+    const port = enabledPort;
     try {
       {
         agentCommand.mockReset();
@@ -416,7 +426,7 @@ describe("OpenAI-compatible HTTP API (e2e)", () => {
         expect(fallbackText).toContain("hello");
       }
     } finally {
-      await server.close({ reason: "test done" });
+      // shared server
     }
   });
 });

src/gateway/openresponses-http.e2e.test.ts

@@ -1,11 +1,23 @@
-import { describe, expect, it } from "vitest";
+import { afterAll, beforeAll, describe, expect, it } from "vitest";
 
 import { HISTORY_CONTEXT_MARKER } from "../auto-reply/reply/history.js";
 import { CURRENT_MESSAGE_MARKER } from "../auto-reply/reply/mentions.js";
 import { emitAgentEvent } from "../infra/agent-events.js";
 import { agentCommand, getFreePort, installGatewayTestHooks } from "./test-helpers.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
+
+let enabledServer: Awaited<ReturnType<typeof startServer>>;
+let enabledPort: number;
+
+beforeAll(async () => {
+  enabledPort = await getFreePort();
+  enabledServer = await startServer(enabledPort);
+});
+
+afterAll(async () => {
+  await enabledServer.close({ reason: "openresponses enabled suite done" });
+});
 
 async function startServerWithDefaultConfig(port: number) {
   const { startGatewayServer } = await import("./server.js");
@@ -72,7 +84,7 @@ async function ensureResponseConsumed(res: Response) {
 describe("OpenResponses HTTP API (e2e)", () => {
   it("rejects when disabled (default + config)", { timeout: 120_000 }, async () => {
     const port = await getFreePort();
-    const server = await startServerWithDefaultConfig(port);
+    const _server = await startServerWithDefaultConfig(port);
     try {
       const res = await postResponses(port, {
         model: "clawdbot",
@@ -81,7 +93,7 @@ describe("OpenResponses HTTP API (e2e)", () => {
       expect(res.status).toBe(404);
       await ensureResponseConsumed(res);
     } finally {
-      await server.close({ reason: "test done" });
+      // shared server
     }
 
     const disabledPort = await getFreePort();
@@ -101,8 +113,7 @@ describe("OpenResponses HTTP API (e2e)", () => {
   });
 
   it("handles OpenResponses request parsing and validation", async () => {
-    const port = await getFreePort();
-    const server = await startServer(port);
+    const port = enabledPort;
     const mockAgentOnce = (payloads: Array<{ text: string }>, meta?: unknown) => {
       agentCommand.mockReset();
       agentCommand.mockResolvedValueOnce({ payloads, meta } as never);
@@ -406,14 +417,12 @@ describe("OpenResponses HTTP API (e2e)", () => {
       );
       await ensureResponseConsumed(resNoUser);
     } finally {
-      await server.close({ reason: "test done" });
+      // shared server
     }
   });
 
   it("streams OpenResponses SSE events", async () => {
-    const port = await getFreePort();
-    const server = await startServer(port);
-
+    const port = enabledPort;
     try {
       agentCommand.mockReset();
       agentCommand.mockImplementationOnce(async (opts: unknown) => {
@@ -489,7 +498,7 @@ describe("OpenResponses HTTP API (e2e)", () => {
         expect(event.event).toBe(parsed.type);
       }
     } finally {
-      await server.close({ reason: "test done" });
+      // shared server
     }
   });
 });

src/gateway/server-chat.ts

@@ -173,7 +173,7 @@ export function createAgentEventHandler({
     nodeSendToSession(sessionKey, "chat", payload);
   };
 
-  const shouldEmitVerboseEvents = (runId: string, sessionKey?: string) => {
+  const shouldEmitToolEvents = (runId: string, sessionKey?: string) => {
     const runContext = getAgentRunContext(runId);
     const runVerbose = normalizeVerboseLevel(runContext?.verboseLevel);
     if (runVerbose) return runVerbose === "on";
@@ -198,10 +198,7 @@ export function createAgentEventHandler({
     // Include sessionKey so Control UI can filter tool streams per session.
     const agentPayload = sessionKey ? { ...evt, sessionKey } : evt;
     const last = agentRunSeq.get(evt.runId) ?? 0;
-    if (
-      (evt.stream === "tool" || evt.stream === "compaction") &&
-      !shouldEmitVerboseEvents(evt.runId, sessionKey)
-    ) {
+    if (evt.stream === "tool" && !shouldEmitToolEvents(evt.runId, sessionKey)) {
       agentRunSeq.set(evt.runId, evt.seq);
       return;
     }

src/gateway/server.agent.gateway-server-agent-a.test.ts

@@ -1,9 +1,10 @@
 import fs from "node:fs/promises";
 import os from "node:os";
 import path from "node:path";
-import { describe, expect, test, vi } from "vitest";
+import { afterAll, beforeAll, describe, expect, test, vi } from "vitest";
 import type { ChannelPlugin } from "../channels/plugins/types.js";
 import type { PluginRegistry } from "../plugins/registry.js";
+import { setActivePluginRegistry } from "../plugins/runtime.js";
 import {
   agentCommand,
   connectOk,
@@ -14,7 +15,22 @@ import {
   writeSessionStore,
 } from "./test-helpers.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
+
+let server: Awaited<ReturnType<typeof startServerWithClient>>["server"];
+let ws: Awaited<ReturnType<typeof startServerWithClient>>["ws"];
+
+beforeAll(async () => {
+  const started = await startServerWithClient();
+  server = started.server;
+  ws = started.ws;
+  await connectOk(ws);
+});
+
+afterAll(async () => {
+  ws.close();
+  await server.close();
+});
 
 const registryState = vi.hoisted(() => ({
   registry: {
@@ -43,6 +59,11 @@ vi.mock("./server-plugins.js", async () => {
   };
 });
 
+const setRegistry = (registry: PluginRegistry) => {
+  registryState.registry = registry;
+  setActivePluginRegistry(registry);
+};
+
 const BASE_IMAGE_PNG =
   "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8/x8AAwMCAO+X3mIAAAAASUVORK5CYII=";
 
@@ -142,7 +163,7 @@ const defaultRegistry = createRegistry([
 
 describe("gateway server agent", () => {
   test("agent marks implicit delivery when lastTo is stale", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     testState.allowFrom = ["+436769770569"];
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
@@ -156,10 +177,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -175,14 +192,11 @@ describe("gateway server agent", () => {
     expect(call.to).toBe("+1555");
     expect(call.deliveryTargetMode).toBe("implicit");
     expect(call.sessionId).toBe("sess-main-stale");
-
-    ws.close();
-    await server.close();
     testState.allowFrom = undefined;
   });
 
   test("agent forwards sessionKey to agentCommand", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
     await writeSessionStore({
@@ -193,10 +207,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "agent:main:subagent:abc",
@@ -211,13 +221,10 @@ describe("gateway server agent", () => {
     expectChannels(call, "webchat");
     expect(call.deliver).toBe(false);
     expect(call.to).toBeUndefined();
-
-    ws.close();
-    await server.close();
   });
 
   test("agent derives sessionKey from agentId", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
     testState.agentsConfig = { list: [{ id: "ops" }] };
@@ -230,10 +237,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       agentId: "ops",
@@ -245,16 +248,10 @@ describe("gateway server agent", () => {
     const call = spy.mock.calls.at(-1)?.[0] as Record<string, unknown>;
     expect(call.sessionKey).toBe("agent:ops:main");
     expect(call.sessionId).toBe("sess-ops");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent rejects unknown reply channel", async () => {
-    registryState.registry = defaultRegistry;
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
+    setRegistry(defaultRegistry);
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       replyChannel: "unknown-channel",
@@ -265,18 +262,11 @@ describe("gateway server agent", () => {
 
     const spy = vi.mocked(agentCommand);
     expect(spy).not.toHaveBeenCalled();
-
-    ws.close();
-    await server.close();
   });
 
   test("agent rejects mismatched agentId and sessionKey", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     testState.agentsConfig = { list: [{ id: "ops" }] };
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       agentId: "ops",
@@ -288,13 +278,10 @@ describe("gateway server agent", () => {
 
     const spy = vi.mocked(agentCommand);
     expect(spy).not.toHaveBeenCalled();
-
-    ws.close();
-    await server.close();
   });
 
   test("agent forwards accountId to agentCommand", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     testState.allowFrom = ["+1555"];
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
@@ -309,10 +296,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -329,14 +312,11 @@ describe("gateway server agent", () => {
     expect(call.accountId).toBe("kev");
     const runContext = call.runContext as { accountId?: string } | undefined;
     expect(runContext?.accountId).toBe("kev");
-
-    ws.close();
-    await server.close();
     testState.allowFrom = undefined;
   });
 
   test("agent avoids lastAccountId when explicit to is provided", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     testState.allowFrom = ["+1555"];
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
@@ -351,10 +331,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -369,14 +345,11 @@ describe("gateway server agent", () => {
     expectChannels(call, "whatsapp");
     expect(call.to).toBe("+1666");
     expect(call.accountId).toBeUndefined();
-
-    ws.close();
-    await server.close();
     testState.allowFrom = undefined;
   });
 
   test("agent keeps explicit accountId when explicit to is provided", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     testState.allowFrom = ["+1555"];
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
@@ -391,10 +364,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -410,14 +379,11 @@ describe("gateway server agent", () => {
     expectChannels(call, "whatsapp");
     expect(call.to).toBe("+1666");
     expect(call.accountId).toBe("primary");
-
-    ws.close();
-    await server.close();
     testState.allowFrom = undefined;
   });
 
   test("agent falls back to lastAccountId for implicit delivery", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     testState.allowFrom = ["+1555"];
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
@@ -432,10 +398,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -449,14 +411,11 @@ describe("gateway server agent", () => {
     expectChannels(call, "whatsapp");
     expect(call.to).toBe("+1555");
     expect(call.accountId).toBe("kev");
-
-    ws.close();
-    await server.close();
     testState.allowFrom = undefined;
   });
 
   test("agent forwards image attachments as images[]", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
     await writeSessionStore({
@@ -467,10 +426,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "what is in the image?",
       sessionKey: "main",
@@ -497,13 +452,10 @@ describe("gateway server agent", () => {
     expect(images[0]?.type).toBe("image");
     expect(images[0]?.mimeType).toBe("image/png");
     expect(images[0]?.data).toBe(BASE_IMAGE_PNG);
-
-    ws.close();
-    await server.close();
   });
 
   test("agent falls back to whatsapp when delivery requested and no last channel exists", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     testState.allowFrom = ["+1555"];
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
@@ -515,10 +467,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -533,14 +481,11 @@ describe("gateway server agent", () => {
     expect(call.to).toBe("+1555");
     expect(call.deliver).toBe(true);
     expect(call.sessionId).toBe("sess-main-missing-provider");
-
-    ws.close();
-    await server.close();
     testState.allowFrom = undefined;
   });
 
   test("agent routes main last-channel whatsapp", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
     await writeSessionStore({
@@ -553,10 +498,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -574,13 +515,10 @@ describe("gateway server agent", () => {
     expect(call.deliver).toBe(true);
     expect(call.bestEffortDeliver).toBe(true);
     expect(call.sessionId).toBe("sess-main-whatsapp");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent routes main last-channel telegram", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
     await writeSessionStore({
@@ -593,10 +531,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -613,13 +547,10 @@ describe("gateway server agent", () => {
     expect(call.deliver).toBe(true);
     expect(call.bestEffortDeliver).toBe(true);
     expect(call.sessionId).toBe("sess-main");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent routes main last-channel discord", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
     await writeSessionStore({
@@ -632,10 +563,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -652,13 +579,10 @@ describe("gateway server agent", () => {
     expect(call.deliver).toBe(true);
     expect(call.bestEffortDeliver).toBe(true);
     expect(call.sessionId).toBe("sess-discord");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent routes main last-channel slack", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
     await writeSessionStore({
@@ -671,10 +595,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -691,13 +611,10 @@ describe("gateway server agent", () => {
     expect(call.deliver).toBe(true);
     expect(call.bestEffortDeliver).toBe(true);
     expect(call.sessionId).toBe("sess-slack");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent routes main last-channel signal", async () => {
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
     testState.sessionStorePath = path.join(dir, "sessions.json");
     await writeSessionStore({
@@ -710,10 +627,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -730,8 +643,5 @@ describe("gateway server agent", () => {
     expect(call.deliver).toBe(true);
     expect(call.bestEffortDeliver).toBe(true);
     expect(call.sessionId).toBe("sess-signal");
-
-    ws.close();
-    await server.close();
   });
 });

src/gateway/server.agent.gateway-server-agent-b.test.ts

@@ -1,7 +1,7 @@
 import fs from "node:fs/promises";
 import os from "node:os";
 import path from "node:path";
-import { afterEach, beforeEach, describe, expect, test, vi } from "vitest";
+import { afterAll, afterEach, beforeAll, beforeEach, describe, expect, test, vi } from "vitest";
 import { WebSocket } from "ws";
 import type { ChannelPlugin } from "../channels/plugins/types.js";
 import { emitAgentEvent, registerAgentRunContext } from "../infra/agent-events.js";
@@ -22,7 +22,24 @@ import {
   writeSessionStore,
 } from "./test-helpers.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
+
+let server: Awaited<ReturnType<typeof startServerWithClient>>["server"];
+let ws: Awaited<ReturnType<typeof startServerWithClient>>["ws"];
+let port: number;
+
+beforeAll(async () => {
+  const started = await startServerWithClient();
+  server = started.server;
+  ws = started.ws;
+  port = started.port;
+  await connectOk(ws);
+});
+
+afterAll(async () => {
+  ws.close();
+  await server.close();
+});
 
 const registryState = vi.hoisted(() => ({
   registry: {
@@ -130,10 +147,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -150,9 +163,6 @@ describe("gateway server agent", () => {
     expect(call.deliver).toBe(true);
     expect(call.bestEffortDeliver).toBe(true);
     expect(call.sessionId).toBe("sess-teams");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent accepts channel aliases (imsg/teams)", async () => {
@@ -177,10 +187,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const resIMessage = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -208,15 +214,9 @@ describe("gateway server agent", () => {
     const lastTeamsCall = spy.mock.calls.at(-1)?.[0] as Record<string, unknown>;
     expectChannels(lastTeamsCall, "msteams");
     expect(lastTeamsCall.to).toBe("conversation:teams-abc");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent rejects unknown channel", async () => {
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -225,9 +225,6 @@ describe("gateway server agent", () => {
     });
     expect(res.ok).toBe(false);
     expect(res.error?.code).toBe("INVALID_REQUEST");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent ignores webchat last-channel for routing", async () => {
@@ -244,10 +241,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -264,9 +257,6 @@ describe("gateway server agent", () => {
     expect(call.deliver).toBe(true);
     expect(call.bestEffortDeliver).toBe(true);
     expect(call.sessionId).toBe("sess-main-webchat");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent uses webchat for internal runs when last provider is webchat", async () => {
@@ -282,10 +272,6 @@ describe("gateway server agent", () => {
         },
       },
     });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "agent", {
       message: "hi",
       sessionKey: "main",
@@ -302,15 +288,9 @@ describe("gateway server agent", () => {
     expect(call.deliver).toBe(false);
     expect(call.bestEffortDeliver).toBe(true);
     expect(call.sessionId).toBe("sess-main-webchat-internal");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent ack response then final response", { timeout: 8000 }, async () => {
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const ackP = onceMessage(
       ws,
       (o) => o.type === "res" && o.id === "ag1" && o.payload?.status === "accepted",
@@ -333,15 +313,9 @@ describe("gateway server agent", () => {
     expect(ack.payload.runId).toBeDefined();
     expect(final.payload.runId).toBe(ack.payload.runId);
     expect(final.payload.status).toBe("ok");
-
-    ws.close();
-    await server.close();
   });
 
   test("agent dedupes by idempotencyKey after completion", async () => {
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const firstFinalP = onceMessage(
       ws,
       (o) => o.type === "res" && o.id === "ag1" && o.payload?.status !== "accepted",
@@ -367,9 +341,6 @@ describe("gateway server agent", () => {
     );
     const second = await secondP;
     expect(second.payload).toEqual(firstFinal.payload);
-
-    ws.close();
-    await server.close();
   });
 
   test("agent dedupe survives reconnect", { timeout: 60_000 }, async () => {
@@ -433,8 +404,9 @@ describe("gateway server agent", () => {
       },
     });
 
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws, {
+    const webchatWs = new WebSocket(`ws://127.0.0.1:${port}`);
+    await new Promise<void>((resolve) => webchatWs.once("open", resolve));
+    await connectOk(webchatWs, {
       client: {
         id: GATEWAY_CLIENT_NAMES.WEBCHAT,
         version: "1.0.0",
@@ -446,7 +418,7 @@ describe("gateway server agent", () => {
     registerAgentRunContext("run-auto-1", { sessionKey: "main" });
 
     const finalChatP = onceMessage(
-      ws,
+      webchatWs,
       (o) => {
         if (o.type !== "event" || o.event !== "chat") return false;
         const payload = o.payload as { state?: unknown; runId?: unknown } | undefined;
@@ -474,7 +446,6 @@ describe("gateway server agent", () => {
     expect(payload.sessionKey).toBe("main");
     expect(payload.runId).toBe("run-auto-1");
 
-    ws.close();
-    await server.close();
+    webchatWs.close();
   });
 });

src/gateway/server.agent.gateway-server-agent.test.ts

@@ -1,218 +0,0 @@
-import fs from "node:fs/promises";
-import os from "node:os";
-import path from "node:path";
-import { describe, expect, test } from "vitest";
-import { emitAgentEvent, registerAgentRunContext } from "../infra/agent-events.js";
-import { GATEWAY_CLIENT_MODES, GATEWAY_CLIENT_NAMES } from "../utils/message-channel.js";
-import {
-  connectOk,
-  installGatewayTestHooks,
-  onceMessage,
-  rpcReq,
-  startServerWithClient,
-  testState,
-  writeSessionStore,
-} from "./test-helpers.js";
-
-installGatewayTestHooks();
-
-const _BASE_IMAGE_PNG =
-  "iVBORw0KGgoAAAANSUhEUgAAAAEAAAABCAQAAAC1HAwCAAAAC0lEQVR42mP8/x8AAwMCAO+X3mIAAAAASUVORK5CYII=";
-
-function _expectChannels(call: Record<string, unknown>, channel: string) {
-  expect(call.channel).toBe(channel);
-  expect(call.messageChannel).toBe(channel);
-}
-
-describe("gateway server agent", () => {
-  test("agent events include sessionKey and agent.wait covers lifecycle flows", async () => {
-    const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
-    testState.sessionStorePath = path.join(dir, "sessions.json");
-    await writeSessionStore({
-      entries: {
-        main: {
-          sessionId: "sess-main",
-          updatedAt: Date.now(),
-          verboseLevel: "off",
-        },
-      },
-    });
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws, {
-      client: {
-        id: GATEWAY_CLIENT_NAMES.WEBCHAT,
-        version: "1.0.0",
-        platform: "test",
-        mode: GATEWAY_CLIENT_MODES.WEBCHAT,
-      },
-    });
-
-    registerAgentRunContext("run-tool-1", {
-      sessionKey: "main",
-      verboseLevel: "on",
-    });
-
-    {
-      const agentEvtP = onceMessage(
-        ws,
-        (o) => o.type === "event" && o.event === "agent" && o.payload?.runId === "run-tool-1",
-        8000,
-      );
-
-      emitAgentEvent({
-        runId: "run-tool-1",
-        stream: "tool",
-        data: { phase: "start", name: "read", toolCallId: "tool-1" },
-      });
-
-      const evt = await agentEvtP;
-      const payload =
-        evt.payload && typeof evt.payload === "object"
-          ? (evt.payload as Record<string, unknown>)
-          : {};
-      expect(payload.sessionKey).toBe("main");
-    }
-
-    {
-      registerAgentRunContext("run-tool-off", { sessionKey: "agent:main:main" });
-
-      const compactionEvtP = onceMessage(
-        ws,
-        (o) =>
-          o.type === "event" &&
-          o.event === "agent" &&
-          o.payload?.runId === "run-tool-off" &&
-          o.payload?.stream === "compaction",
-        1000,
-      );
-
-      emitAgentEvent({
-        runId: "run-tool-off",
-        stream: "compaction",
-        data: { phase: "start" },
-      });
-
-      emitAgentEvent({
-        runId: "run-tool-off",
-        stream: "tool",
-        data: { phase: "start", name: "read", toolCallId: "tool-1" },
-      });
-      emitAgentEvent({
-        runId: "run-tool-off",
-        stream: "assistant",
-        data: { text: "hello" },
-      });
-
-      const evt = await onceMessage(
-        ws,
-        (o) => o.type === "event" && o.event === "agent" && o.payload?.runId === "run-tool-off",
-        8000,
-      );
-      const payload =
-        evt.payload && typeof evt.payload === "object"
-          ? (evt.payload as Record<string, unknown>)
-          : {};
-      expect(payload.stream).toBe("assistant");
-
-      await expect(compactionEvtP).rejects.toThrow("timeout");
-    }
-
-    {
-      const waitP = rpcReq(ws, "agent.wait", {
-        runId: "run-wait-1",
-        timeoutMs: 1000,
-      });
-
-      setTimeout(() => {
-        emitAgentEvent({
-          runId: "run-wait-1",
-          stream: "lifecycle",
-          data: { phase: "end", startedAt: 200, endedAt: 210 },
-        });
-      }, 5);
-
-      const res = await waitP;
-      expect(res.ok).toBe(true);
-      expect(res.payload.status).toBe("ok");
-      expect(res.payload.startedAt).toBe(200);
-    }
-
-    {
-      emitAgentEvent({
-        runId: "run-wait-early",
-        stream: "lifecycle",
-        data: { phase: "end", startedAt: 50, endedAt: 55 },
-      });
-
-      const res = await rpcReq(ws, "agent.wait", {
-        runId: "run-wait-early",
-        timeoutMs: 1000,
-      });
-      expect(res.ok).toBe(true);
-      expect(res.payload.status).toBe("ok");
-      expect(res.payload.startedAt).toBe(50);
-    }
-
-    {
-      const res = await rpcReq(ws, "agent.wait", {
-        runId: "run-wait-3",
-        timeoutMs: 30,
-      });
-      expect(res.ok).toBe(true);
-      expect(res.payload.status).toBe("timeout");
-    }
-
-    {
-      const waitP = rpcReq(ws, "agent.wait", {
-        runId: "run-wait-err",
-        timeoutMs: 1000,
-      });
-
-      setTimeout(() => {
-        emitAgentEvent({
-          runId: "run-wait-err",
-          stream: "lifecycle",
-          data: { phase: "error", error: "boom" },
-        });
-      }, 5);
-
-      const res = await waitP;
-      expect(res.ok).toBe(true);
-      expect(res.payload.status).toBe("error");
-      expect(res.payload.error).toBe("boom");
-    }
-
-    {
-      const waitP = rpcReq(ws, "agent.wait", {
-        runId: "run-wait-start",
-        timeoutMs: 1000,
-      });
-
-      emitAgentEvent({
-        runId: "run-wait-start",
-        stream: "lifecycle",
-        data: { phase: "start", startedAt: 123 },
-      });
-
-      setTimeout(() => {
-        emitAgentEvent({
-          runId: "run-wait-start",
-          stream: "lifecycle",
-          data: { phase: "end", endedAt: 456 },
-        });
-      }, 5);
-
-      const res = await waitP;
-      expect(res.ok).toBe(true);
-      expect(res.payload.status).toBe("ok");
-      expect(res.payload.startedAt).toBe(123);
-      expect(res.payload.endedAt).toBe(456);
-    }
-
-    ws.close();
-    await server.close();
-    await fs.rm(dir, { recursive: true, force: true });
-    testState.sessionStorePath = undefined;
-  });
-});

src/gateway/server.agents.test.ts

@@ -1,44 +0,0 @@
-import { describe, expect, test } from "vitest";
-import {
-  connectOk,
-  installGatewayTestHooks,
-  rpcReq,
-  startServerWithClient,
-  testState,
-} from "./test-helpers.js";
-
-installGatewayTestHooks();
-
-describe("gateway server agents", () => {
-  test("lists configured agents via agents.list RPC", async () => {
-    testState.agentsConfig = {
-      list: [
-        { id: "work", name: "Work", default: true },
-        { id: "home", name: "Home" },
-      ],
-    };
-
-    const { ws } = await startServerWithClient();
-    const hello = await connectOk(ws);
-    expect((hello as unknown as { features?: { methods?: string[] } }).features?.methods).toEqual(
-      expect.arrayContaining(["agents.list"]),
-    );
-
-    const res = await rpcReq<{
-      defaultId: string;
-      mainKey: string;
-      scope: string;
-      agents: Array<{ id: string; name?: string }>;
-    }>(ws, "agents.list", {});
-
-    expect(res.ok).toBe(true);
-    expect(res.payload?.defaultId).toBe("work");
-    expect(res.payload?.mainKey).toBe("main");
-    expect(res.payload?.scope).toBe("per-sender");
-    expect(res.payload?.agents.map((agent) => agent.id)).toEqual(["work", "home", "main"]);
-    const work = res.payload?.agents.find((agent) => agent.id === "work");
-    const home = res.payload?.agents.find((agent) => agent.id === "home");
-    expect(work?.name).toBe("Work");
-    expect(home?.name).toBe("Home");
-  });
-});

src/gateway/server.auth.test.ts

@@ -13,7 +13,7 @@ import {
 } from "./test-helpers.js";
 import { GATEWAY_CLIENT_MODES, GATEWAY_CLIENT_NAMES } from "../utils/message-channel.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
 
 async function waitForWsClose(ws: WebSocket, timeoutMs: number): Promise<boolean> {
   if (ws.readyState === WebSocket.CLOSED) return true;

src/gateway/server.channels.test.ts

@@ -1,6 +1,7 @@
-import { afterEach, describe, expect, test, vi } from "vitest";
+import { afterAll, beforeAll, describe, expect, test, vi } from "vitest";
 import type { ChannelPlugin } from "../channels/plugins/types.js";
 import type { PluginRegistry } from "../plugins/registry.js";
+import { setActivePluginRegistry } from "../plugins/runtime.js";
 import {
   connectOk,
   installGatewayTestHooks,
@@ -10,7 +11,7 @@ import {
 
 const loadConfigHelpers = async () => await import("../config/config.js");
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
 
 const registryState = vi.hoisted(() => ({
   registry: {
@@ -131,30 +132,31 @@ const defaultRegistry = createRegistry([
   },
 ]);
 
-const servers: Array<Awaited<ReturnType<typeof startServerWithClient>>> = [];
+let server: Awaited<ReturnType<typeof startServerWithClient>>["server"];
+let ws: Awaited<ReturnType<typeof startServerWithClient>>["ws"];
 
-afterEach(async () => {
-  for (const { server, ws } of servers) {
-    try {
-      ws.close();
-      await server.close();
-    } catch {
-      /* ignore */
-    }
-  }
-  servers.length = 0;
-  await new Promise((resolve) => setTimeout(resolve, 50));
+beforeAll(async () => {
+  setRegistry(defaultRegistry);
+  const started = await startServerWithClient();
+  server = started.server;
+  ws = started.ws;
+  await connectOk(ws);
 });
 
+afterAll(async () => {
+  ws.close();
+  await server.close();
+});
+
+function setRegistry(registry: PluginRegistry) {
+  registryState.registry = registry;
+  setActivePluginRegistry(registry);
+}
+
 describe("gateway server channels", () => {
   test("channels.status returns snapshot without probe", async () => {
     vi.stubEnv("TELEGRAM_BOT_TOKEN", undefined);
-    registryState.registry = defaultRegistry;
-    const result = await startServerWithClient();
-    servers.push(result);
-    const { ws } = result;
-    await connectOk(ws);
-
+    setRegistry(defaultRegistry);
     const res = await rpcReq<{
       channels?: Record<
         string,
@@ -181,12 +183,7 @@ describe("gateway server channels", () => {
   });
 
   test("channels.logout reports no session when missing", async () => {
-    registryState.registry = defaultRegistry;
-    const result = await startServerWithClient();
-    servers.push(result);
-    const { ws } = result;
-    await connectOk(ws);
-
+    setRegistry(defaultRegistry);
     const res = await rpcReq<{ cleared?: boolean; channel?: string }>(ws, "channels.logout", {
       channel: "whatsapp",
     });
@@ -197,7 +194,7 @@ describe("gateway server channels", () => {
 
   test("channels.logout clears telegram bot token from config", async () => {
     vi.stubEnv("TELEGRAM_BOT_TOKEN", undefined);
-    registryState.registry = defaultRegistry;
+    setRegistry(defaultRegistry);
     const { readConfigFileSnapshot, writeConfigFile } = await loadConfigHelpers();
     await writeConfigFile({
       channels: {
@@ -207,12 +204,6 @@ describe("gateway server channels", () => {
         },
       },
     });
-
-    const result = await startServerWithClient();
-    servers.push(result);
-    const { ws } = result;
-    await connectOk(ws);
-
     const res = await rpcReq<{
       cleared?: boolean;
       envToken?: boolean;

src/gateway/server.chat.gateway-server-chat-b.test.ts

@@ -14,7 +14,7 @@ import {
   testState,
   writeSessionStore,
 } from "./test-helpers.js";
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
 async function waitFor(condition: () => boolean, timeoutMs = 1500) {
   const deadline = Date.now() + timeoutMs;
   while (Date.now() < deadline) {
@@ -38,355 +38,88 @@ const sendReq = (
     }),
   );
 };
-const withSessionStore = async <T>(
-  tempDirs: string[],
-  entries: Record<
-    string,
-    { sessionId: string; updatedAt: number; lastChannel?: string; lastTo?: string }
-  >,
-  fn: (dir: string) => Promise<T>,
-): Promise<T> => {
-  const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
-  tempDirs.push(dir);
-  testState.sessionStorePath = path.join(dir, "sessions.json");
-  await writeSessionStore({ entries });
-  try {
-    return await fn(dir);
-  } finally {
-    testState.sessionStorePath = undefined;
-  }
-};
 describe("gateway server chat", () => {
-  test("handles history, abort, idempotency, and ordering flows", { timeout: 60_000 }, async () => {
-    const tempDirs: string[] = [];
-    const { server, ws } = await startServerWithClient();
-    const spy = vi.mocked(agentCommand);
-    const resetSpy = () => {
-      spy.mockReset();
-      spy.mockResolvedValue(undefined);
-    };
-    try {
-      await connectOk(ws);
-      await withSessionStore(
-        tempDirs,
-        { main: { sessionId: "sess-main", updatedAt: Date.now() } },
-        async (historyDir) => {
-          const bigText = "x".repeat(200_000);
-          const largeLines: string[] = [];
-          for (let i = 0; i < 40; i += 1) {
-            largeLines.push(
-              JSON.stringify({
-                message: {
-                  role: "user",
-                  content: [{ type: "text", text: `${i}:${bigText}` }],
-                  timestamp: Date.now() + i,
-                },
-              }),
-            );
-          }
-          await fs.writeFile(
-            path.join(historyDir, "sess-main.jsonl"),
-            largeLines.join("\n"),
-            "utf-8",
+  const timeoutMs = process.platform === "win32" ? 120_000 : 60_000;
+  test(
+    "handles history, abort, idempotency, and ordering flows",
+    { timeout: timeoutMs },
+    async () => {
+      const tempDirs: string[] = [];
+      const { server, ws } = await startServerWithClient();
+      const spy = vi.mocked(agentCommand);
+      const resetSpy = () => {
+        spy.mockReset();
+        spy.mockResolvedValue(undefined);
+      };
+      try {
+        await connectOk(ws);
+        const sessionDir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
+        tempDirs.push(sessionDir);
+        testState.sessionStorePath = path.join(sessionDir, "sessions.json");
+        const writeStore = async (
+          entries: Record<
+            string,
+            { sessionId: string; updatedAt: number; lastChannel?: string; lastTo?: string }
+          >,
+        ) => {
+          await writeSessionStore({ entries });
+        };
+
+        await writeStore({ main: { sessionId: "sess-main", updatedAt: Date.now() } });
+        const bigText = "x".repeat(155_000);
+        const largeLines: string[] = [];
+        for (let i = 0; i < 40; i += 1) {
+          largeLines.push(
+            JSON.stringify({
+              message: {
+                role: "user",
+                content: [{ type: "text", text: `${i}:${bigText}` }],
+                timestamp: Date.now() + i,
+              },
+            }),
           );
-          const cappedRes = await rpcReq<{ messages?: unknown[] }>(ws, "chat.history", {
-            sessionKey: "main",
-            limit: 1000,
-          });
-          expect(cappedRes.ok).toBe(true);
-          const cappedMsgs = cappedRes.payload?.messages ?? [];
-          const bytes = Buffer.byteLength(JSON.stringify(cappedMsgs), "utf8");
-          expect(bytes).toBeLessThanOrEqual(6 * 1024 * 1024);
-          expect(cappedMsgs.length).toBeLessThan(60);
-        },
-      );
-      await withSessionStore(
-        tempDirs,
-        {
+        }
+        await fs.writeFile(
+          path.join(sessionDir, "sess-main.jsonl"),
+          largeLines.join("\n"),
+          "utf-8",
+        );
+        const cappedRes = await rpcReq<{ messages?: unknown[] }>(ws, "chat.history", {
+          sessionKey: "main",
+          limit: 1000,
+        });
+        expect(cappedRes.ok).toBe(true);
+        const cappedMsgs = cappedRes.payload?.messages ?? [];
+        const bytes = Buffer.byteLength(JSON.stringify(cappedMsgs), "utf8");
+        expect(bytes).toBeLessThanOrEqual(6 * 1024 * 1024);
+        expect(cappedMsgs.length).toBeLessThan(60);
+
+        await writeStore({
           main: {
             sessionId: "sess-main",
             updatedAt: Date.now(),
             lastChannel: "whatsapp",
             lastTo: "+1555",
           },
-        },
-        async () => {
-          const routeRes = await rpcReq(ws, "chat.send", {
-            sessionKey: "main",
-            message: "hello",
-            idempotencyKey: "idem-route",
-          });
-          expect(routeRes.ok).toBe(true);
-          const stored = JSON.parse(
-            await fs.readFile(testState.sessionStorePath as string, "utf-8"),
-          ) as Record<string, { lastChannel?: string; lastTo?: string } | undefined>;
-          expect(stored["agent:main:main"]?.lastChannel).toBe("whatsapp");
-          expect(stored["agent:main:main"]?.lastTo).toBe("+1555");
-        },
-      );
-      await withSessionStore(
-        tempDirs,
-        { main: { sessionId: "sess-main", updatedAt: Date.now() } },
-        async () => {
-          resetSpy();
-          let abortInFlight: Promise<unknown> | undefined;
-          try {
-            const callsBefore = spy.mock.calls.length;
-            spy.mockImplementationOnce(async (opts) => {
-              const signal = (opts as { abortSignal?: AbortSignal }).abortSignal;
-              await new Promise<void>((resolve) => {
-                if (!signal) return resolve();
-                if (signal.aborted) return resolve();
-                signal.addEventListener("abort", () => resolve(), { once: true });
-              });
-            });
-            const sendResP = onceMessage(
-              ws,
-              (o) => o.type === "res" && o.id === "send-abort-1",
-              8000,
-            );
-            const abortResP = onceMessage(ws, (o) => o.type === "res" && o.id === "abort-1", 8000);
-            const abortedEventP = onceMessage(
-              ws,
-              (o) => o.type === "event" && o.event === "chat" && o.payload?.state === "aborted",
-              8000,
-            );
-            abortInFlight = Promise.allSettled([sendResP, abortResP, abortedEventP]);
-            sendReq(ws, "send-abort-1", "chat.send", {
-              sessionKey: "main",
-              message: "hello",
-              idempotencyKey: "idem-abort-1",
-              timeoutMs: 30_000,
-            });
-            const sendRes = await sendResP;
-            expect(sendRes.ok).toBe(true);
-            await new Promise<void>((resolve, reject) => {
-              const deadline = Date.now() + 1000;
-              const tick = () => {
-                if (spy.mock.calls.length > callsBefore) return resolve();
-                if (Date.now() > deadline)
-                  return reject(new Error("timeout waiting for agentCommand"));
-                setTimeout(tick, 5);
-              };
-              tick();
-            });
-            sendReq(ws, "abort-1", "chat.abort", {
-              sessionKey: "main",
-              runId: "idem-abort-1",
-            });
-            const abortRes = await abortResP;
-            expect(abortRes.ok).toBe(true);
-            const evt = await abortedEventP;
-            expect(evt.payload?.runId).toBe("idem-abort-1");
-            expect(evt.payload?.sessionKey).toBe("main");
-          } finally {
-            await abortInFlight;
-          }
-        },
-      );
-      await withSessionStore(
-        tempDirs,
-        { main: { sessionId: "sess-main", updatedAt: Date.now() } },
-        async () => {
-          sessionStoreSaveDelayMs.value = 120;
-          resetSpy();
-          try {
-            spy.mockImplementationOnce(async (opts) => {
-              const signal = (opts as { abortSignal?: AbortSignal }).abortSignal;
-              await new Promise<void>((resolve) => {
-                if (!signal) return resolve();
-                if (signal.aborted) return resolve();
-                signal.addEventListener("abort", () => resolve(), { once: true });
-              });
-            });
-            const abortedEventP = onceMessage(
-              ws,
-              (o) => o.type === "event" && o.event === "chat" && o.payload?.state === "aborted",
-            );
-            const sendResP = onceMessage(
-              ws,
-              (o) => o.type === "res" && o.id === "send-abort-save-1",
-            );
-            sendReq(ws, "send-abort-save-1", "chat.send", {
-              sessionKey: "main",
-              message: "hello",
-              idempotencyKey: "idem-abort-save-1",
-              timeoutMs: 30_000,
-            });
-            const abortResP = onceMessage(ws, (o) => o.type === "res" && o.id === "abort-save-1");
-            sendReq(ws, "abort-save-1", "chat.abort", {
-              sessionKey: "main",
-              runId: "idem-abort-save-1",
-            });
-            const abortRes = await abortResP;
-            expect(abortRes.ok).toBe(true);
-            const sendRes = await sendResP;
-            expect(sendRes.ok).toBe(true);
-            const evt = await abortedEventP;
-            expect(evt.payload?.runId).toBe("idem-abort-save-1");
-            expect(evt.payload?.sessionKey).toBe("main");
-          } finally {
-            sessionStoreSaveDelayMs.value = 0;
-          }
-        },
-      );
-      await withSessionStore(
-        tempDirs,
-        { main: { sessionId: "sess-main", updatedAt: Date.now() } },
-        async () => {
-          resetSpy();
-          const callsBeforeStop = spy.mock.calls.length;
-          spy.mockImplementationOnce(async (opts) => {
-            const signal = (opts as { abortSignal?: AbortSignal }).abortSignal;
-            await new Promise<void>((resolve) => {
-              if (!signal) return resolve();
-              if (signal.aborted) return resolve();
-              signal.addEventListener("abort", () => resolve(), { once: true });
-            });
-          });
-          const stopSendResP = onceMessage(
-            ws,
-            (o) => o.type === "res" && o.id === "send-stop-1",
-            8000,
-          );
-          sendReq(ws, "send-stop-1", "chat.send", {
-            sessionKey: "main",
-            message: "hello",
-            idempotencyKey: "idem-stop-run",
-          });
-          const stopSendRes = await stopSendResP;
-          expect(stopSendRes.ok).toBe(true);
-          await waitFor(() => spy.mock.calls.length > callsBeforeStop);
-          const abortedStopEventP = onceMessage(
-            ws,
-            (o) =>
-              o.type === "event" &&
-              o.event === "chat" &&
-              o.payload?.state === "aborted" &&
-              o.payload?.runId === "idem-stop-run",
-            8000,
-          );
-          const stopResP = onceMessage(ws, (o) => o.type === "res" && o.id === "send-stop-2", 8000);
-          sendReq(ws, "send-stop-2", "chat.send", {
-            sessionKey: "main",
-            message: "/stop",
-            idempotencyKey: "idem-stop-req",
-          });
-          const stopRes = await stopResP;
-          expect(stopRes.ok).toBe(true);
-          const stopEvt = await abortedStopEventP;
-          expect(stopEvt.payload?.sessionKey).toBe("main");
-          expect(spy.mock.calls.length).toBe(callsBeforeStop + 1);
-        },
-      );
-      resetSpy();
-      let resolveRun: (() => void) | undefined;
-      const runDone = new Promise<void>((resolve) => {
-        resolveRun = resolve;
-      });
-      spy.mockImplementationOnce(async () => {
-        await runDone;
-      });
-      const started = await rpcReq<{ runId?: string; status?: string }>(ws, "chat.send", {
-        sessionKey: "main",
-        message: "hello",
-        idempotencyKey: "idem-status-1",
-      });
-      expect(started.ok).toBe(true);
-      expect(started.payload?.status).toBe("started");
-      const inFlightRes = await rpcReq<{ runId?: string; status?: string }>(ws, "chat.send", {
-        sessionKey: "main",
-        message: "hello",
-        idempotencyKey: "idem-status-1",
-      });
-      expect(inFlightRes.ok).toBe(true);
-      expect(inFlightRes.payload?.status).toBe("in_flight");
-      resolveRun?.();
-      let completed = false;
-      for (let i = 0; i < 50; i++) {
-        const again = await rpcReq<{ runId?: string; status?: string }>(ws, "chat.send", {
+        });
+        const routeRes = await rpcReq(ws, "chat.send", {
           sessionKey: "main",
           message: "hello",
-          idempotencyKey: "idem-status-1",
+          idempotencyKey: "idem-route",
         });
-        if (again.ok && again.payload?.status === "ok") {
-          completed = true;
-          break;
-        }
-        await new Promise((r) => setTimeout(r, 10));
-      }
-      expect(completed).toBe(true);
-      resetSpy();
-      spy.mockImplementationOnce(async (opts) => {
-        const signal = (opts as { abortSignal?: AbortSignal }).abortSignal;
-        await new Promise<void>((resolve) => {
-          if (!signal) return resolve();
-          if (signal.aborted) return resolve();
-          signal.addEventListener("abort", () => resolve(), { once: true });
-        });
-      });
-      const abortedEventP = onceMessage(
-        ws,
-        (o) =>
-          o.type === "event" &&
-          o.event === "chat" &&
-          o.payload?.state === "aborted" &&
-          o.payload?.runId === "idem-abort-all-1",
-      );
-      const startedAbortAll = await rpcReq(ws, "chat.send", {
-        sessionKey: "main",
-        message: "hello",
-        idempotencyKey: "idem-abort-all-1",
-      });
-      expect(startedAbortAll.ok).toBe(true);
-      const abortRes = await rpcReq<{
-        ok?: boolean;
-        aborted?: boolean;
-        runIds?: string[];
-      }>(ws, "chat.abort", { sessionKey: "main" });
-      expect(abortRes.ok).toBe(true);
-      expect(abortRes.payload?.aborted).toBe(true);
-      expect(abortRes.payload?.runIds ?? []).toContain("idem-abort-all-1");
-      await abortedEventP;
-      const noDeltaP = onceMessage(
-        ws,
-        (o) =>
-          o.type === "event" &&
-          o.event === "chat" &&
-          (o.payload?.state === "delta" || o.payload?.state === "final") &&
-          o.payload?.runId === "idem-abort-all-1",
-        250,
-      );
-      emitAgentEvent({
-        runId: "idem-abort-all-1",
-        stream: "assistant",
-        data: { text: "should be suppressed" },
-      });
-      emitAgentEvent({
-        runId: "idem-abort-all-1",
-        stream: "lifecycle",
-        data: { phase: "end" },
-      });
-      await expect(noDeltaP).rejects.toThrow(/timeout/i);
-      await withSessionStore(tempDirs, {}, async () => {
-        const abortUnknown = await rpcReq<{
-          ok?: boolean;
-          aborted?: boolean;
-        }>(ws, "chat.abort", { sessionKey: "main", runId: "missing-run" });
-        expect(abortUnknown.ok).toBe(true);
-        expect(abortUnknown.payload?.aborted).toBe(false);
-      });
-      await withSessionStore(
-        tempDirs,
-        { main: { sessionId: "sess-main", updatedAt: Date.now() } },
-        async () => {
-          resetSpy();
-          let agentStartedResolve: (() => void) | undefined;
-          const agentStartedP = new Promise<void>((resolve) => {
-            agentStartedResolve = resolve;
-          });
+        expect(routeRes.ok).toBe(true);
+        const stored = JSON.parse(
+          await fs.readFile(testState.sessionStorePath as string, "utf-8"),
+        ) as Record<string, { lastChannel?: string; lastTo?: string } | undefined>;
+        expect(stored["agent:main:main"]?.lastChannel).toBe("whatsapp");
+        expect(stored["agent:main:main"]?.lastTo).toBe("+1555");
+
+        await writeStore({ main: { sessionId: "sess-main", updatedAt: Date.now() } });
+        resetSpy();
+        let abortInFlight: Promise<unknown> | undefined;
+        try {
+          const callsBefore = spy.mock.calls.length;
           spy.mockImplementationOnce(async (opts) => {
-            agentStartedResolve?.();
             const signal = (opts as { abortSignal?: AbortSignal }).abortSignal;
             await new Promise<void>((resolve) => {
               if (!signal) return resolve();
@@ -396,127 +129,356 @@ describe("gateway server chat", () => {
           });
           const sendResP = onceMessage(
             ws,
-            (o) => o.type === "res" && o.id === "send-mismatch-1",
-            10_000,
+            (o) => o.type === "res" && o.id === "send-abort-1",
+            8000,
           );
-          sendReq(ws, "send-mismatch-1", "chat.send", {
+          const abortResP = onceMessage(ws, (o) => o.type === "res" && o.id === "abort-1", 8000);
+          const abortedEventP = onceMessage(
+            ws,
+            (o) => o.type === "event" && o.event === "chat" && o.payload?.state === "aborted",
+            8000,
+          );
+          abortInFlight = Promise.allSettled([sendResP, abortResP, abortedEventP]);
+          sendReq(ws, "send-abort-1", "chat.send", {
             sessionKey: "main",
             message: "hello",
-            idempotencyKey: "idem-mismatch-1",
+            idempotencyKey: "idem-abort-1",
             timeoutMs: 30_000,
           });
-          await agentStartedP;
-          const abortMismatch = await rpcReq(ws, "chat.abort", {
-            sessionKey: "other",
-            runId: "idem-mismatch-1",
-          });
-          expect(abortMismatch.ok).toBe(false);
-          expect(abortMismatch.error?.code).toBe("INVALID_REQUEST");
-          const abortMismatch2 = await rpcReq(ws, "chat.abort", {
-            sessionKey: "main",
-            runId: "idem-mismatch-1",
-          });
-          expect(abortMismatch2.ok).toBe(true);
           const sendRes = await sendResP;
           expect(sendRes.ok).toBe(true);
-        },
-      );
-      await withSessionStore(
-        tempDirs,
-        { main: { sessionId: "sess-main", updatedAt: Date.now() } },
-        async () => {
-          resetSpy();
-          spy.mockResolvedValueOnce(undefined);
-          sendReq(ws, "send-complete-1", "chat.send", {
+          await new Promise<void>((resolve, reject) => {
+            const deadline = Date.now() + 1000;
+            const tick = () => {
+              if (spy.mock.calls.length > callsBefore) return resolve();
+              if (Date.now() > deadline)
+                return reject(new Error("timeout waiting for agentCommand"));
+              setTimeout(tick, 5);
+            };
+            tick();
+          });
+          sendReq(ws, "abort-1", "chat.abort", {
+            sessionKey: "main",
+            runId: "idem-abort-1",
+          });
+          const abortRes = await abortResP;
+          expect(abortRes.ok).toBe(true);
+          const evt = await abortedEventP;
+          expect(evt.payload?.runId).toBe("idem-abort-1");
+          expect(evt.payload?.sessionKey).toBe("main");
+        } finally {
+          await abortInFlight;
+        }
+
+        await writeStore({ main: { sessionId: "sess-main", updatedAt: Date.now() } });
+        sessionStoreSaveDelayMs.value = 120;
+        resetSpy();
+        try {
+          spy.mockImplementationOnce(async (opts) => {
+            const signal = (opts as { abortSignal?: AbortSignal }).abortSignal;
+            await new Promise<void>((resolve) => {
+              if (!signal) return resolve();
+              if (signal.aborted) return resolve();
+              signal.addEventListener("abort", () => resolve(), { once: true });
+            });
+          });
+          const abortedEventP = onceMessage(
+            ws,
+            (o) => o.type === "event" && o.event === "chat" && o.payload?.state === "aborted",
+          );
+          const sendResP = onceMessage(ws, (o) => o.type === "res" && o.id === "send-abort-save-1");
+          sendReq(ws, "send-abort-save-1", "chat.send", {
+            sessionKey: "main",
+            message: "hello",
+            idempotencyKey: "idem-abort-save-1",
+            timeoutMs: 30_000,
+          });
+          const abortResP = onceMessage(ws, (o) => o.type === "res" && o.id === "abort-save-1");
+          sendReq(ws, "abort-save-1", "chat.abort", {
+            sessionKey: "main",
+            runId: "idem-abort-save-1",
+          });
+          const abortRes = await abortResP;
+          expect(abortRes.ok).toBe(true);
+          const sendRes = await sendResP;
+          expect(sendRes.ok).toBe(true);
+          const evt = await abortedEventP;
+          expect(evt.payload?.runId).toBe("idem-abort-save-1");
+          expect(evt.payload?.sessionKey).toBe("main");
+        } finally {
+          sessionStoreSaveDelayMs.value = 0;
+        }
+
+        await writeStore({ main: { sessionId: "sess-main", updatedAt: Date.now() } });
+        resetSpy();
+        const callsBeforeStop = spy.mock.calls.length;
+        spy.mockImplementationOnce(async (opts) => {
+          const signal = (opts as { abortSignal?: AbortSignal }).abortSignal;
+          await new Promise<void>((resolve) => {
+            if (!signal) return resolve();
+            if (signal.aborted) return resolve();
+            signal.addEventListener("abort", () => resolve(), { once: true });
+          });
+        });
+        const stopSendResP = onceMessage(
+          ws,
+          (o) => o.type === "res" && o.id === "send-stop-1",
+          8000,
+        );
+        sendReq(ws, "send-stop-1", "chat.send", {
+          sessionKey: "main",
+          message: "hello",
+          idempotencyKey: "idem-stop-run",
+        });
+        const stopSendRes = await stopSendResP;
+        expect(stopSendRes.ok).toBe(true);
+        await waitFor(() => spy.mock.calls.length > callsBeforeStop);
+        const abortedStopEventP = onceMessage(
+          ws,
+          (o) =>
+            o.type === "event" &&
+            o.event === "chat" &&
+            o.payload?.state === "aborted" &&
+            o.payload?.runId === "idem-stop-run",
+          8000,
+        );
+        const stopResP = onceMessage(ws, (o) => o.type === "res" && o.id === "send-stop-2", 8000);
+        sendReq(ws, "send-stop-2", "chat.send", {
+          sessionKey: "main",
+          message: "/stop",
+          idempotencyKey: "idem-stop-req",
+        });
+        const stopRes = await stopResP;
+        expect(stopRes.ok).toBe(true);
+        const stopEvt = await abortedStopEventP;
+        expect(stopEvt.payload?.sessionKey).toBe("main");
+        expect(spy.mock.calls.length).toBe(callsBeforeStop + 1);
+        resetSpy();
+        let resolveRun: (() => void) | undefined;
+        const runDone = new Promise<void>((resolve) => {
+          resolveRun = resolve;
+        });
+        spy.mockImplementationOnce(async () => {
+          await runDone;
+        });
+        const started = await rpcReq<{ runId?: string; status?: string }>(ws, "chat.send", {
+          sessionKey: "main",
+          message: "hello",
+          idempotencyKey: "idem-status-1",
+        });
+        expect(started.ok).toBe(true);
+        expect(started.payload?.status).toBe("started");
+        const inFlightRes = await rpcReq<{ runId?: string; status?: string }>(ws, "chat.send", {
+          sessionKey: "main",
+          message: "hello",
+          idempotencyKey: "idem-status-1",
+        });
+        expect(inFlightRes.ok).toBe(true);
+        expect(inFlightRes.payload?.status).toBe("in_flight");
+        resolveRun?.();
+        let completed = false;
+        for (let i = 0; i < 20; i++) {
+          const again = await rpcReq<{ runId?: string; status?: string }>(ws, "chat.send", {
+            sessionKey: "main",
+            message: "hello",
+            idempotencyKey: "idem-status-1",
+          });
+          if (again.ok && again.payload?.status === "ok") {
+            completed = true;
+            break;
+          }
+          await new Promise((r) => setTimeout(r, 10));
+        }
+        expect(completed).toBe(true);
+        resetSpy();
+        spy.mockImplementationOnce(async (opts) => {
+          const signal = (opts as { abortSignal?: AbortSignal }).abortSignal;
+          await new Promise<void>((resolve) => {
+            if (!signal) return resolve();
+            if (signal.aborted) return resolve();
+            signal.addEventListener("abort", () => resolve(), { once: true });
+          });
+        });
+        const abortedEventP = onceMessage(
+          ws,
+          (o) =>
+            o.type === "event" &&
+            o.event === "chat" &&
+            o.payload?.state === "aborted" &&
+            o.payload?.runId === "idem-abort-all-1",
+        );
+        const startedAbortAll = await rpcReq(ws, "chat.send", {
+          sessionKey: "main",
+          message: "hello",
+          idempotencyKey: "idem-abort-all-1",
+        });
+        expect(startedAbortAll.ok).toBe(true);
+        const abortRes = await rpcReq<{
+          ok?: boolean;
+          aborted?: boolean;
+          runIds?: string[];
+        }>(ws, "chat.abort", { sessionKey: "main" });
+        expect(abortRes.ok).toBe(true);
+        expect(abortRes.payload?.aborted).toBe(true);
+        expect(abortRes.payload?.runIds ?? []).toContain("idem-abort-all-1");
+        await abortedEventP;
+        const noDeltaP = onceMessage(
+          ws,
+          (o) =>
+            o.type === "event" &&
+            o.event === "chat" &&
+            (o.payload?.state === "delta" || o.payload?.state === "final") &&
+            o.payload?.runId === "idem-abort-all-1",
+          250,
+        );
+        emitAgentEvent({
+          runId: "idem-abort-all-1",
+          stream: "assistant",
+          data: { text: "should be suppressed" },
+        });
+        emitAgentEvent({
+          runId: "idem-abort-all-1",
+          stream: "lifecycle",
+          data: { phase: "end" },
+        });
+        await expect(noDeltaP).rejects.toThrow(/timeout/i);
+        await writeStore({});
+        const abortUnknown = await rpcReq<{
+          ok?: boolean;
+          aborted?: boolean;
+        }>(ws, "chat.abort", { sessionKey: "main", runId: "missing-run" });
+        expect(abortUnknown.ok).toBe(true);
+        expect(abortUnknown.payload?.aborted).toBe(false);
+
+        await writeStore({ main: { sessionId: "sess-main", updatedAt: Date.now() } });
+        resetSpy();
+        let agentStartedResolve: (() => void) | undefined;
+        const agentStartedP = new Promise<void>((resolve) => {
+          agentStartedResolve = resolve;
+        });
+        spy.mockImplementationOnce(async (opts) => {
+          agentStartedResolve?.();
+          const signal = (opts as { abortSignal?: AbortSignal }).abortSignal;
+          await new Promise<void>((resolve) => {
+            if (!signal) return resolve();
+            if (signal.aborted) return resolve();
+            signal.addEventListener("abort", () => resolve(), { once: true });
+          });
+        });
+        const sendResP = onceMessage(
+          ws,
+          (o) => o.type === "res" && o.id === "send-mismatch-1",
+          10_000,
+        );
+        sendReq(ws, "send-mismatch-1", "chat.send", {
+          sessionKey: "main",
+          message: "hello",
+          idempotencyKey: "idem-mismatch-1",
+          timeoutMs: 30_000,
+        });
+        await agentStartedP;
+        const abortMismatch = await rpcReq(ws, "chat.abort", {
+          sessionKey: "other",
+          runId: "idem-mismatch-1",
+        });
+        expect(abortMismatch.ok).toBe(false);
+        expect(abortMismatch.error?.code).toBe("INVALID_REQUEST");
+        const abortMismatch2 = await rpcReq(ws, "chat.abort", {
+          sessionKey: "main",
+          runId: "idem-mismatch-1",
+        });
+        expect(abortMismatch2.ok).toBe(true);
+        const sendRes = await sendResP;
+        expect(sendRes.ok).toBe(true);
+
+        await writeStore({ main: { sessionId: "sess-main", updatedAt: Date.now() } });
+        resetSpy();
+        spy.mockResolvedValueOnce(undefined);
+        sendReq(ws, "send-complete-1", "chat.send", {
+          sessionKey: "main",
+          message: "hello",
+          idempotencyKey: "idem-complete-1",
+          timeoutMs: 30_000,
+        });
+        const sendCompleteRes = await onceMessage(
+          ws,
+          (o) => o.type === "res" && o.id === "send-complete-1",
+        );
+        expect(sendCompleteRes.ok).toBe(true);
+        let completedRun = false;
+        for (let i = 0; i < 20; i++) {
+          const again = await rpcReq<{ runId?: string; status?: string }>(ws, "chat.send", {
             sessionKey: "main",
             message: "hello",
             idempotencyKey: "idem-complete-1",
             timeoutMs: 30_000,
           });
-          const sendCompleteRes = await onceMessage(
-            ws,
-            (o) => o.type === "res" && o.id === "send-complete-1",
-          );
-          expect(sendCompleteRes.ok).toBe(true);
-          let completedRun = false;
-          for (let i = 0; i < 50; i++) {
-            const again = await rpcReq<{ runId?: string; status?: string }>(ws, "chat.send", {
-              sessionKey: "main",
-              message: "hello",
-              idempotencyKey: "idem-complete-1",
-              timeoutMs: 30_000,
-            });
-            if (again.ok && again.payload?.status === "ok") {
-              completedRun = true;
-              break;
-            }
-            await new Promise((r) => setTimeout(r, 10));
+          if (again.ok && again.payload?.status === "ok") {
+            completedRun = true;
+            break;
           }
-          expect(completedRun).toBe(true);
-          const abortCompleteRes = await rpcReq(ws, "chat.abort", {
-            sessionKey: "main",
-            runId: "idem-complete-1",
-          });
-          expect(abortCompleteRes.ok).toBe(true);
-          expect(abortCompleteRes.payload?.aborted).toBe(false);
-        },
-      );
-      await withSessionStore(
-        tempDirs,
-        { main: { sessionId: "sess-main", updatedAt: Date.now() } },
-        async () => {
-          const res1 = await rpcReq(ws, "chat.send", {
-            sessionKey: "main",
-            message: "first",
-            idempotencyKey: "idem-1",
-          });
-          expect(res1.ok).toBe(true);
-          const res2 = await rpcReq(ws, "chat.send", {
-            sessionKey: "main",
-            message: "second",
-            idempotencyKey: "idem-2",
-          });
-          expect(res2.ok).toBe(true);
-          const final1P = onceMessage(
-            ws,
-            (o) => o.type === "event" && o.event === "chat" && o.payload?.state === "final",
-            8000,
-          );
-          emitAgentEvent({
-            runId: "idem-1",
-            stream: "lifecycle",
-            data: { phase: "end" },
-          });
-          const final1 = await final1P;
-          const run1 =
-            final1.payload && typeof final1.payload === "object"
-              ? (final1.payload as { runId?: string }).runId
-              : undefined;
-          expect(run1).toBe("idem-1");
-          const final2P = onceMessage(
-            ws,
-            (o) => o.type === "event" && o.event === "chat" && o.payload?.state === "final",
-            8000,
-          );
-          emitAgentEvent({
-            runId: "idem-2",
-            stream: "lifecycle",
-            data: { phase: "end" },
-          });
-          const final2 = await final2P;
-          const run2 =
-            final2.payload && typeof final2.payload === "object"
-              ? (final2.payload as { runId?: string }).runId
-              : undefined;
-          expect(run2).toBe("idem-2");
-        },
-      );
-    } finally {
-      testState.sessionStorePath = undefined;
-      sessionStoreSaveDelayMs.value = 0;
-      ws.close();
-      await server.close();
-      await Promise.all(tempDirs.map((dir) => fs.rm(dir, { recursive: true, force: true })));
-    }
-  });
+          await new Promise((r) => setTimeout(r, 10));
+        }
+        expect(completedRun).toBe(true);
+        const abortCompleteRes = await rpcReq(ws, "chat.abort", {
+          sessionKey: "main",
+          runId: "idem-complete-1",
+        });
+        expect(abortCompleteRes.ok).toBe(true);
+        expect(abortCompleteRes.payload?.aborted).toBe(false);
+
+        await writeStore({ main: { sessionId: "sess-main", updatedAt: Date.now() } });
+        const res1 = await rpcReq(ws, "chat.send", {
+          sessionKey: "main",
+          message: "first",
+          idempotencyKey: "idem-1",
+        });
+        expect(res1.ok).toBe(true);
+        const res2 = await rpcReq(ws, "chat.send", {
+          sessionKey: "main",
+          message: "second",
+          idempotencyKey: "idem-2",
+        });
+        expect(res2.ok).toBe(true);
+        const final1P = onceMessage(
+          ws,
+          (o) => o.type === "event" && o.event === "chat" && o.payload?.state === "final",
+          8000,
+        );
+        emitAgentEvent({
+          runId: "idem-1",
+          stream: "lifecycle",
+          data: { phase: "end" },
+        });
+        const final1 = await final1P;
+        const run1 =
+          final1.payload && typeof final1.payload === "object"
+            ? (final1.payload as { runId?: string }).runId
+            : undefined;
+        expect(run1).toBe("idem-1");
+        const final2P = onceMessage(
+          ws,
+          (o) => o.type === "event" && o.event === "chat" && o.payload?.state === "final",
+          8000,
+        );
+        emitAgentEvent({
+          runId: "idem-2",
+          stream: "lifecycle",
+          data: { phase: "end" },
+        });
+        const final2 = await final2P;
+        const run2 =
+          final2.payload && typeof final2.payload === "object"
+            ? (final2.payload as { runId?: string }).runId
+            : undefined;
+        expect(run2).toBe("idem-2");
+      } finally {
+        testState.sessionStorePath = undefined;
+        sessionStoreSaveDelayMs.value = 0;
+        ws.close();
+        await server.close();
+        await Promise.all(tempDirs.map((dir) => fs.rm(dir, { recursive: true, force: true })));
+      }
+    },
+  );
 });

src/gateway/server.chat.gateway-server-chat.test.ts

@@ -1,9 +1,10 @@
 import fs from "node:fs/promises";
 import os from "node:os";
 import path from "node:path";
-import { describe, expect, test, vi } from "vitest";
+import { afterAll, beforeAll, describe, expect, test, vi } from "vitest";
 import { WebSocket } from "ws";
 import { GATEWAY_CLIENT_MODES, GATEWAY_CLIENT_NAMES } from "../utils/message-channel.js";
+import { emitAgentEvent, registerAgentRunContext } from "../infra/agent-events.js";
 import {
   agentCommand,
   connectOk,
@@ -15,7 +16,24 @@ import {
   writeSessionStore,
 } from "./test-helpers.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
+
+let server: Awaited<ReturnType<typeof startServerWithClient>>["server"];
+let ws: WebSocket;
+let port: number;
+
+beforeAll(async () => {
+  const started = await startServerWithClient();
+  server = started.server;
+  ws = started.ws;
+  port = started.port;
+  await connectOk(ws);
+});
+
+afterAll(async () => {
+  ws.close();
+  await server.close();
+});
 
 async function waitFor(condition: () => boolean, timeoutMs = 1500) {
   const deadline = Date.now() + timeoutMs;
@@ -29,12 +47,9 @@ async function waitFor(condition: () => boolean, timeoutMs = 1500) {
 describe("gateway server chat", () => {
   test("handles chat send and history flows", async () => {
     const tempDirs: string[] = [];
-    const { server, ws, port } = await startServerWithClient();
     let webchatWs: WebSocket | undefined;
 
     try {
-      await connectOk(ws);
-
       webchatWs = new WebSocket(`ws://127.0.0.1:${port}`);
       await new Promise<void>((resolve) => webchatWs?.once("open", resolve));
       await connectOk(webchatWs, {
@@ -240,9 +255,182 @@ describe("gateway server chat", () => {
       testState.sessionStorePath = undefined;
       testState.sessionConfig = undefined;
       if (webchatWs) webchatWs.close();
-      ws.close();
-      await server.close();
       await Promise.all(tempDirs.map((dir) => fs.rm(dir, { recursive: true, force: true })));
     }
   });
+
+  test("agent events include sessionKey and agent.wait covers lifecycle flows", async () => {
+    const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-gw-"));
+    testState.sessionStorePath = path.join(dir, "sessions.json");
+    await writeSessionStore({
+      entries: {
+        main: {
+          sessionId: "sess-main",
+          updatedAt: Date.now(),
+          verboseLevel: "off",
+        },
+      },
+    });
+
+    const webchatWs = new WebSocket(`ws://127.0.0.1:${port}`);
+    await new Promise<void>((resolve) => webchatWs.once("open", resolve));
+    await connectOk(webchatWs, {
+      client: {
+        id: GATEWAY_CLIENT_NAMES.WEBCHAT,
+        version: "1.0.0",
+        platform: "test",
+        mode: GATEWAY_CLIENT_MODES.WEBCHAT,
+      },
+    });
+
+    try {
+      registerAgentRunContext("run-tool-1", {
+        sessionKey: "main",
+        verboseLevel: "on",
+      });
+
+      {
+        const agentEvtP = onceMessage(
+          webchatWs,
+          (o) => o.type === "event" && o.event === "agent" && o.payload?.runId === "run-tool-1",
+          8000,
+        );
+
+        emitAgentEvent({
+          runId: "run-tool-1",
+          stream: "tool",
+          data: { phase: "start", name: "read", toolCallId: "tool-1" },
+        });
+
+        const evt = await agentEvtP;
+        const payload =
+          evt.payload && typeof evt.payload === "object"
+            ? (evt.payload as Record<string, unknown>)
+            : {};
+        expect(payload.sessionKey).toBe("main");
+      }
+
+      {
+        registerAgentRunContext("run-tool-off", { sessionKey: "agent:main:main" });
+
+        emitAgentEvent({
+          runId: "run-tool-off",
+          stream: "tool",
+          data: { phase: "start", name: "read", toolCallId: "tool-1" },
+        });
+        emitAgentEvent({
+          runId: "run-tool-off",
+          stream: "assistant",
+          data: { text: "hello" },
+        });
+
+        const evt = await onceMessage(
+          webchatWs,
+          (o) => o.type === "event" && o.event === "agent" && o.payload?.runId === "run-tool-off",
+          8000,
+        );
+        const payload =
+          evt.payload && typeof evt.payload === "object"
+            ? (evt.payload as Record<string, unknown>)
+            : {};
+        expect(payload.stream).toBe("assistant");
+      }
+
+      {
+        const waitP = rpcReq(webchatWs, "agent.wait", {
+          runId: "run-wait-1",
+          timeoutMs: 1000,
+        });
+
+        setTimeout(() => {
+          emitAgentEvent({
+            runId: "run-wait-1",
+            stream: "lifecycle",
+            data: { phase: "end", startedAt: 200, endedAt: 210 },
+          });
+        }, 5);
+
+        const res = await waitP;
+        expect(res.ok).toBe(true);
+        expect(res.payload.status).toBe("ok");
+        expect(res.payload.startedAt).toBe(200);
+      }
+
+      {
+        emitAgentEvent({
+          runId: "run-wait-early",
+          stream: "lifecycle",
+          data: { phase: "end", startedAt: 50, endedAt: 55 },
+        });
+
+        const res = await rpcReq(webchatWs, "agent.wait", {
+          runId: "run-wait-early",
+          timeoutMs: 1000,
+        });
+        expect(res.ok).toBe(true);
+        expect(res.payload.status).toBe("ok");
+        expect(res.payload.startedAt).toBe(50);
+      }
+
+      {
+        const res = await rpcReq(webchatWs, "agent.wait", {
+          runId: "run-wait-3",
+          timeoutMs: 30,
+        });
+        expect(res.ok).toBe(true);
+        expect(res.payload.status).toBe("timeout");
+      }
+
+      {
+        const waitP = rpcReq(webchatWs, "agent.wait", {
+          runId: "run-wait-err",
+          timeoutMs: 1000,
+        });
+
+        setTimeout(() => {
+          emitAgentEvent({
+            runId: "run-wait-err",
+            stream: "lifecycle",
+            data: { phase: "error", error: "boom" },
+          });
+        }, 5);
+
+        const res = await waitP;
+        expect(res.ok).toBe(true);
+        expect(res.payload.status).toBe("error");
+        expect(res.payload.error).toBe("boom");
+      }
+
+      {
+        const waitP = rpcReq(webchatWs, "agent.wait", {
+          runId: "run-wait-start",
+          timeoutMs: 1000,
+        });
+
+        emitAgentEvent({
+          runId: "run-wait-start",
+          stream: "lifecycle",
+          data: { phase: "start", startedAt: 123 },
+        });
+
+        setTimeout(() => {
+          emitAgentEvent({
+            runId: "run-wait-start",
+            stream: "lifecycle",
+            data: { phase: "end", endedAt: 456 },
+          });
+        }, 5);
+
+        const res = await waitP;
+        expect(res.ok).toBe(true);
+        expect(res.payload.status).toBe("ok");
+        expect(res.payload.startedAt).toBe(123);
+        expect(res.payload.endedAt).toBe(456);
+      }
+    } finally {
+      webchatWs.close();
+      await fs.rm(dir, { recursive: true, force: true });
+      testState.sessionStorePath = undefined;
+    }
+  });
 });

src/gateway/server.config-patch.test.ts

@@ -1,4 +1,7 @@
-import { describe, expect, it } from "vitest";
+import fs from "node:fs/promises";
+import os from "node:os";
+import path from "node:path";
+import { afterAll, beforeAll, describe, expect, it } from "vitest";
 
 import { resolveConfigSnapshotHash } from "../config/config.js";
 
@@ -6,16 +9,31 @@ import {
   connectOk,
   installGatewayTestHooks,
   onceMessage,
+  rpcReq,
   startServerWithClient,
+  testState,
+  writeSessionStore,
 } from "./test-helpers.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
+
+let server: Awaited<ReturnType<typeof startServerWithClient>>["server"];
+let ws: Awaited<ReturnType<typeof startServerWithClient>>["ws"];
+
+beforeAll(async () => {
+  const started = await startServerWithClient();
+  server = started.server;
+  ws = started.ws;
+  await connectOk(ws);
+});
+
+afterAll(async () => {
+  ws.close();
+  await server.close();
+});
 
 describe("gateway config.patch", () => {
   it("merges patches without clobbering unrelated config", async () => {
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const setId = "req-set";
     ws.send(
       JSON.stringify({
@@ -100,15 +118,9 @@ describe("gateway config.patch", () => {
     expect(get2Res.ok).toBe(true);
     expect(get2Res.payload?.config?.gateway?.mode).toBe("local");
     expect(get2Res.payload?.config?.channels?.telegram?.botToken).toBe("token-1");
-
-    ws.close();
-    await server.close();
   });
 
   it("requires base hash when config exists", async () => {
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const setId = "req-set-2";
     ws.send(
       JSON.stringify({
@@ -145,15 +157,9 @@ describe("gateway config.patch", () => {
     );
     expect(patchRes.ok).toBe(false);
     expect(patchRes.error?.message).toContain("base hash");
-
-    ws.close();
-    await server.close();
   });
 
   it("requires base hash for config.set when config exists", async () => {
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const setId = "req-set-3";
     ws.send(
       JSON.stringify({
@@ -192,8 +198,108 @@ describe("gateway config.patch", () => {
     );
     expect(set2Res.ok).toBe(false);
     expect(set2Res.error?.message).toContain("base hash");
+  });
+});
 
-    ws.close();
-    await server.close();
+describe("gateway server sessions", () => {
+  it("filters sessions by agentId", async () => {
+    const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-sessions-agents-"));
+    testState.sessionConfig = {
+      store: path.join(dir, "{agentId}", "sessions.json"),
+    };
+    testState.agentsConfig = {
+      list: [{ id: "home", default: true }, { id: "work" }],
+    };
+    const homeDir = path.join(dir, "home");
+    const workDir = path.join(dir, "work");
+    await fs.mkdir(homeDir, { recursive: true });
+    await fs.mkdir(workDir, { recursive: true });
+    await writeSessionStore({
+      storePath: path.join(homeDir, "sessions.json"),
+      agentId: "home",
+      entries: {
+        main: {
+          sessionId: "sess-home-main",
+          updatedAt: Date.now(),
+        },
+        "discord:group:dev": {
+          sessionId: "sess-home-group",
+          updatedAt: Date.now() - 1000,
+        },
+      },
+    });
+    await writeSessionStore({
+      storePath: path.join(workDir, "sessions.json"),
+      agentId: "work",
+      entries: {
+        main: {
+          sessionId: "sess-work-main",
+          updatedAt: Date.now(),
+        },
+      },
+    });
+
+    const homeSessions = await rpcReq<{
+      sessions: Array<{ key: string }>;
+    }>(ws, "sessions.list", {
+      includeGlobal: false,
+      includeUnknown: false,
+      agentId: "home",
+    });
+    expect(homeSessions.ok).toBe(true);
+    expect(homeSessions.payload?.sessions.map((s) => s.key).sort()).toEqual([
+      "agent:home:discord:group:dev",
+      "agent:home:main",
+    ]);
+
+    const workSessions = await rpcReq<{
+      sessions: Array<{ key: string }>;
+    }>(ws, "sessions.list", {
+      includeGlobal: false,
+      includeUnknown: false,
+      agentId: "work",
+    });
+    expect(workSessions.ok).toBe(true);
+    expect(workSessions.payload?.sessions.map((s) => s.key)).toEqual(["agent:work:main"]);
+  });
+
+  it("resolves and patches main alias to default agent main key", async () => {
+    const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-sessions-"));
+    const storePath = path.join(dir, "sessions.json");
+    testState.sessionStorePath = storePath;
+    testState.agentsConfig = { list: [{ id: "ops", default: true }] };
+    testState.sessionConfig = { mainKey: "work" };
+
+    await writeSessionStore({
+      storePath,
+      agentId: "ops",
+      mainKey: "work",
+      entries: {
+        main: {
+          sessionId: "sess-ops-main",
+          updatedAt: Date.now(),
+        },
+      },
+    });
+
+    const resolved = await rpcReq<{ ok: true; key: string }>(ws, "sessions.resolve", {
+      key: "main",
+    });
+    expect(resolved.ok).toBe(true);
+    expect(resolved.payload?.key).toBe("agent:ops:work");
+
+    const patched = await rpcReq<{ ok: true; key: string }>(ws, "sessions.patch", {
+      key: "main",
+      thinkingLevel: "medium",
+    });
+    expect(patched.ok).toBe(true);
+    expect(patched.payload?.key).toBe("agent:ops:work");
+
+    const stored = JSON.parse(await fs.readFile(storePath, "utf-8")) as Record<
+      string,
+      { thinkingLevel?: string }
+    >;
+    expect(stored["agent:ops:work"]?.thinkingLevel).toBe("medium");
+    expect(stored.main).toBeUndefined();
   });
 });

src/gateway/server.cron.test.ts

@@ -1,7 +1,7 @@
 import fs from "node:fs/promises";
 import os from "node:os";
 import path from "node:path";
-import { describe, expect, test, vi } from "vitest";
+import { describe, expect, test } from "vitest";
 import {
   connectOk,
   installGatewayTestHooks,
@@ -36,7 +36,11 @@ async function rmTempDir(dir: string) {
   await fs.rm(dir, { recursive: true, force: true });
 }
 
-async function waitForCronFinished(ws: { send: (data: string) => void }, jobId: string) {
+async function waitForCronFinished(
+  ws: { send: (data: string) => void },
+  jobId: string,
+  timeoutMs = 20_000,
+) {
   await onceMessage(
     ws as never,
     (o) =>
@@ -44,7 +48,7 @@ async function waitForCronFinished(ws: { send: (data: string) => void }, jobId:
       o.event === "cron" &&
       o.payload?.action === "finished" &&
       o.payload?.jobId === jobId,
-    10_000,
+    timeoutMs,
   );
 }
 
@@ -345,14 +349,7 @@ describe("gateway server cron", () => {
       const autoJobId = typeof autoJobIdValue === "string" ? autoJobIdValue : "";
       expect(autoJobId.length > 0).toBe(true);
 
-      vi.useFakeTimers();
-      try {
-        const autoFinishedP = waitForCronFinished(ws, autoJobId);
-        await vi.advanceTimersByTimeAsync(1000);
-        await autoFinishedP;
-      } finally {
-        vi.useRealTimers();
-      }
+      await waitForCronFinished(ws, autoJobId);
 
       await waitForNonEmptyFile(path.join(dir, "cron", "runs", `${autoJobId}.jsonl`));
       const autoEntries = (await rpcReq(ws, "cron.runs", { id: autoJobId, limit: 10 })).payload as

src/gateway/server.hooks.test.ts

@@ -10,206 +10,150 @@ import {
   waitForSystemEvent,
 } from "./test-helpers.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
 
 const resolveMainKey = () => resolveMainSessionKeyFromConfig();
 
 describe("gateway server hooks", () => {
-  test("hooks wake requires auth", async () => {
+  test("handles auth, wake, and agent flows", async () => {
     testState.hooksConfig = { enabled: true, token: "hook-secret" };
     const port = await getFreePort();
     const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({ text: "Ping" }),
-    });
-    expect(res.status).toBe(401);
-    await server.close();
-  });
+    try {
+      const resNoAuth = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ text: "Ping" }),
+      });
+      expect(resNoAuth.status).toBe(401);
 
-  test("hooks wake enqueues system event", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        Authorization: "Bearer hook-secret",
-      },
-      body: JSON.stringify({ text: "Ping", mode: "next-heartbeat" }),
-    });
-    expect(res.status).toBe(200);
-    const events = await waitForSystemEvent();
-    expect(events.some((e) => e.includes("Ping"))).toBe(true);
-    drainSystemEvents(resolveMainKey());
-    await server.close();
-  });
+      const resWake = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          Authorization: "Bearer hook-secret",
+        },
+        body: JSON.stringify({ text: "Ping", mode: "next-heartbeat" }),
+      });
+      expect(resWake.status).toBe(200);
+      const wakeEvents = await waitForSystemEvent();
+      expect(wakeEvents.some((e) => e.includes("Ping"))).toBe(true);
+      drainSystemEvents(resolveMainKey());
 
-  test("hooks agent posts summary to main", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    cronIsolatedRun.mockResolvedValueOnce({
-      status: "ok",
-      summary: "done",
-    });
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/agent`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        Authorization: "Bearer hook-secret",
-      },
-      body: JSON.stringify({ message: "Do it", name: "Email" }),
-    });
-    expect(res.status).toBe(202);
-    const events = await waitForSystemEvent();
-    expect(events.some((e) => e.includes("Hook Email: done"))).toBe(true);
-    drainSystemEvents(resolveMainKey());
-    await server.close();
-  });
+      cronIsolatedRun.mockReset();
+      cronIsolatedRun.mockResolvedValueOnce({
+        status: "ok",
+        summary: "done",
+      });
+      const resAgent = await fetch(`http://127.0.0.1:${port}/hooks/agent`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          Authorization: "Bearer hook-secret",
+        },
+        body: JSON.stringify({ message: "Do it", name: "Email" }),
+      });
+      expect(resAgent.status).toBe(202);
+      const agentEvents = await waitForSystemEvent();
+      expect(agentEvents.some((e) => e.includes("Hook Email: done"))).toBe(true);
+      drainSystemEvents(resolveMainKey());
 
-  test("hooks agent forwards model override", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    cronIsolatedRun.mockClear();
-    cronIsolatedRun.mockResolvedValueOnce({
-      status: "ok",
-      summary: "done",
-    });
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/agent`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        Authorization: "Bearer hook-secret",
-      },
-      body: JSON.stringify({
-        message: "Do it",
-        name: "Email",
-        model: "openai/gpt-4.1-mini",
-      }),
-    });
-    expect(res.status).toBe(202);
-    await waitForSystemEvent();
-    const call = cronIsolatedRun.mock.calls[0]?.[0] as {
-      job?: { payload?: { model?: string } };
-    };
-    expect(call?.job?.payload?.model).toBe("openai/gpt-4.1-mini");
-    drainSystemEvents(resolveMainKey());
-    await server.close();
-  });
+      cronIsolatedRun.mockReset();
+      cronIsolatedRun.mockResolvedValueOnce({
+        status: "ok",
+        summary: "done",
+      });
+      const resAgentModel = await fetch(`http://127.0.0.1:${port}/hooks/agent`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          Authorization: "Bearer hook-secret",
+        },
+        body: JSON.stringify({
+          message: "Do it",
+          name: "Email",
+          model: "openai/gpt-4.1-mini",
+        }),
+      });
+      expect(resAgentModel.status).toBe(202);
+      await waitForSystemEvent();
+      const call = cronIsolatedRun.mock.calls[0]?.[0] as {
+        job?: { payload?: { model?: string } };
+      };
+      expect(call?.job?.payload?.model).toBe("openai/gpt-4.1-mini");
+      drainSystemEvents(resolveMainKey());
 
-  test("hooks wake accepts query token", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/wake?token=hook-secret`, {
-      method: "POST",
-      headers: { "Content-Type": "application/json" },
-      body: JSON.stringify({ text: "Query auth" }),
-    });
-    expect(res.status).toBe(200);
-    const events = await waitForSystemEvent();
-    expect(events.some((e) => e.includes("Query auth"))).toBe(true);
-    drainSystemEvents(resolveMainKey());
-    await server.close();
-  });
+      const resQuery = await fetch(`http://127.0.0.1:${port}/hooks/wake?token=hook-secret`, {
+        method: "POST",
+        headers: { "Content-Type": "application/json" },
+        body: JSON.stringify({ text: "Query auth" }),
+      });
+      expect(resQuery.status).toBe(200);
+      const queryEvents = await waitForSystemEvent();
+      expect(queryEvents.some((e) => e.includes("Query auth"))).toBe(true);
+      drainSystemEvents(resolveMainKey());
 
-  test("hooks agent rejects invalid channel", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/agent`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        Authorization: "Bearer hook-secret",
-      },
-      body: JSON.stringify({ message: "Nope", channel: "sms" }),
-    });
-    expect(res.status).toBe(400);
-    expect(peekSystemEvents(resolveMainKey()).length).toBe(0);
-    await server.close();
-  });
+      const resBadChannel = await fetch(`http://127.0.0.1:${port}/hooks/agent`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          Authorization: "Bearer hook-secret",
+        },
+        body: JSON.stringify({ message: "Nope", channel: "sms" }),
+      });
+      expect(resBadChannel.status).toBe(400);
+      expect(peekSystemEvents(resolveMainKey()).length).toBe(0);
 
-  test("hooks wake accepts x-clawdbot-token header", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        "x-clawdbot-token": "hook-secret",
-      },
-      body: JSON.stringify({ text: "Header auth" }),
-    });
-    expect(res.status).toBe(200);
-    const events = await waitForSystemEvent();
-    expect(events.some((e) => e.includes("Header auth"))).toBe(true);
-    drainSystemEvents(resolveMainKey());
-    await server.close();
-  });
+      const resHeader = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          "x-clawdbot-token": "hook-secret",
+        },
+        body: JSON.stringify({ text: "Header auth" }),
+      });
+      expect(resHeader.status).toBe(200);
+      const headerEvents = await waitForSystemEvent();
+      expect(headerEvents.some((e) => e.includes("Header auth"))).toBe(true);
+      drainSystemEvents(resolveMainKey());
 
-  test("hooks rejects non-post", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
-      method: "GET",
-      headers: { Authorization: "Bearer hook-secret" },
-    });
-    expect(res.status).toBe(405);
-    await server.close();
-  });
+      const resGet = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
+        method: "GET",
+        headers: { Authorization: "Bearer hook-secret" },
+      });
+      expect(resGet.status).toBe(405);
 
-  test("hooks wake requires text", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        Authorization: "Bearer hook-secret",
-      },
-      body: JSON.stringify({ text: " " }),
-    });
-    expect(res.status).toBe(400);
-    await server.close();
-  });
+      const resBlankText = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          Authorization: "Bearer hook-secret",
+        },
+        body: JSON.stringify({ text: " " }),
+      });
+      expect(resBlankText.status).toBe(400);
 
-  test("hooks agent requires message", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/agent`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        Authorization: "Bearer hook-secret",
-      },
-      body: JSON.stringify({ message: " " }),
-    });
-    expect(res.status).toBe(400);
-    await server.close();
-  });
+      const resBlankMessage = await fetch(`http://127.0.0.1:${port}/hooks/agent`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          Authorization: "Bearer hook-secret",
+        },
+        body: JSON.stringify({ message: " " }),
+      });
+      expect(resBlankMessage.status).toBe(400);
 
-  test("hooks rejects invalid json", async () => {
-    testState.hooksConfig = { enabled: true, token: "hook-secret" };
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    const res = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
-      method: "POST",
-      headers: {
-        "Content-Type": "application/json",
-        Authorization: "Bearer hook-secret",
-      },
-      body: "{",
-    });
-    expect(res.status).toBe(400);
-    await server.close();
+      const resBadJson = await fetch(`http://127.0.0.1:${port}/hooks/wake`, {
+        method: "POST",
+        headers: {
+          "Content-Type": "application/json",
+          Authorization: "Bearer hook-secret",
+        },
+        body: "{",
+      });
+      expect(resBadJson.status).toBe(400);
+    } finally {
+      await server.close();
+    }
   });
 });

src/gateway/server.misc.test.ts

@@ -1,189 +0,0 @@
-import fs from "node:fs/promises";
-import { createServer } from "node:net";
-import path from "node:path";
-import { describe, expect, test } from "vitest";
-import { resolveCanvasHostUrl } from "../infra/canvas-host-url.js";
-import { getChannelPlugin } from "../channels/plugins/index.js";
-import { GatewayLockError } from "../infra/gateway-lock.js";
-import type { ChannelOutboundAdapter } from "../channels/plugins/types.js";
-import type { PluginRegistry } from "../plugins/registry.js";
-import { getActivePluginRegistry, setActivePluginRegistry } from "../plugins/runtime.js";
-import { createOutboundTestPlugin } from "../test-utils/channel-plugins.js";
-import {
-  connectOk,
-  getFreePort,
-  installGatewayTestHooks,
-  occupyPort,
-  onceMessage,
-  startGatewayServer,
-  startServerWithClient,
-  testState,
-  testTailnetIPv4,
-} from "./test-helpers.js";
-
-installGatewayTestHooks();
-
-const whatsappOutbound: ChannelOutboundAdapter = {
-  deliveryMode: "direct",
-  sendText: async ({ deps, to, text }) => {
-    if (!deps?.sendWhatsApp) {
-      throw new Error("Missing sendWhatsApp dep");
-    }
-    return { channel: "whatsapp", ...(await deps.sendWhatsApp(to, text, {})) };
-  },
-  sendMedia: async ({ deps, to, text, mediaUrl }) => {
-    if (!deps?.sendWhatsApp) {
-      throw new Error("Missing sendWhatsApp dep");
-    }
-    return { channel: "whatsapp", ...(await deps.sendWhatsApp(to, text, { mediaUrl })) };
-  },
-};
-
-const whatsappPlugin = createOutboundTestPlugin({
-  id: "whatsapp",
-  outbound: whatsappOutbound,
-  label: "WhatsApp",
-});
-
-const createRegistry = (channels: PluginRegistry["channels"]): PluginRegistry => ({
-  plugins: [],
-  tools: [],
-  channels,
-  providers: [],
-  gatewayHandlers: {},
-  httpHandlers: [],
-  cliRegistrars: [],
-  services: [],
-  diagnostics: [],
-});
-
-const whatsappRegistry = createRegistry([
-  {
-    pluginId: "whatsapp",
-    source: "test",
-    plugin: whatsappPlugin,
-  },
-]);
-const emptyRegistry = createRegistry([]);
-
-describe("gateway server misc", () => {
-  test("hello-ok advertises the gateway port for canvas host", async () => {
-    const prevToken = process.env.CLAWDBOT_GATEWAY_TOKEN;
-    const prevCanvasPort = process.env.CLAWDBOT_CANVAS_HOST_PORT;
-    process.env.CLAWDBOT_GATEWAY_TOKEN = "secret";
-    testTailnetIPv4.value = "100.64.0.1";
-    testState.gatewayBind = "lan";
-    const canvasPort = await getFreePort();
-    testState.canvasHostPort = canvasPort;
-    process.env.CLAWDBOT_CANVAS_HOST_PORT = String(canvasPort);
-
-    const port = await getFreePort();
-    const canvasHostUrl = resolveCanvasHostUrl({
-      canvasPort,
-      requestHost: `100.64.0.1:${port}`,
-      localAddress: "127.0.0.1",
-    });
-    expect(canvasHostUrl).toBe(`http://100.64.0.1:${canvasPort}`);
-    if (prevToken === undefined) {
-      delete process.env.CLAWDBOT_GATEWAY_TOKEN;
-    } else {
-      process.env.CLAWDBOT_GATEWAY_TOKEN = prevToken;
-    }
-    if (prevCanvasPort === undefined) {
-      delete process.env.CLAWDBOT_CANVAS_HOST_PORT;
-    } else {
-      process.env.CLAWDBOT_CANVAS_HOST_PORT = prevCanvasPort;
-    }
-  });
-
-  test("send dedupes by idempotencyKey", { timeout: 60_000 }, async () => {
-    const prevRegistry = getActivePluginRegistry() ?? emptyRegistry;
-    try {
-      const { server, ws } = await startServerWithClient();
-      await connectOk(ws);
-      setActivePluginRegistry(whatsappRegistry);
-      expect(getChannelPlugin("whatsapp")).toBeDefined();
-
-      const idem = "same-key";
-      const res1P = onceMessage(ws, (o) => o.type === "res" && o.id === "a1");
-      const res2P = onceMessage(ws, (o) => o.type === "res" && o.id === "a2");
-      const sendReq = (id: string) =>
-        ws.send(
-          JSON.stringify({
-            type: "req",
-            id,
-            method: "send",
-            params: { to: "+15550000000", message: "hi", idempotencyKey: idem },
-          }),
-        );
-      sendReq("a1");
-      sendReq("a2");
-
-      const res1 = await res1P;
-      const res2 = await res2P;
-      expect(res1.ok).toBe(true);
-      expect(res2.ok).toBe(true);
-      expect(res1.payload).toEqual(res2.payload);
-      ws.close();
-      await server.close();
-    } finally {
-      setActivePluginRegistry(prevRegistry);
-    }
-  });
-
-  test("auto-enables configured channel plugins on startup", async () => {
-    const configPath = process.env.CLAWDBOT_CONFIG_PATH;
-    if (!configPath) throw new Error("Missing CLAWDBOT_CONFIG_PATH");
-    await fs.mkdir(path.dirname(configPath), { recursive: true });
-    await fs.writeFile(
-      configPath,
-      JSON.stringify(
-        {
-          channels: {
-            discord: {
-              token: "token-123",
-            },
-          },
-        },
-        null,
-        2,
-      ),
-      "utf-8",
-    );
-
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    await server.close();
-
-    const updated = JSON.parse(await fs.readFile(configPath, "utf-8")) as Record<string, unknown>;
-    const plugins = updated.plugins as Record<string, unknown> | undefined;
-    const entries = plugins?.entries as Record<string, unknown> | undefined;
-    const discord = entries?.discord as Record<string, unknown> | undefined;
-    expect(discord?.enabled).toBe(true);
-    expect((updated.channels as Record<string, unknown> | undefined)?.discord).toMatchObject({
-      token: "token-123",
-    });
-  });
-
-  test("refuses to start when port already bound", async () => {
-    const { server: blocker, port } = await occupyPort();
-    await expect(startGatewayServer(port)).rejects.toBeInstanceOf(GatewayLockError);
-    await expect(startGatewayServer(port)).rejects.toThrow(/already listening/i);
-    blocker.close();
-  });
-
-  test("releases port after close", async () => {
-    const port = await getFreePort();
-    const server = await startGatewayServer(port);
-    await server.close();
-
-    const probe = createServer();
-    await new Promise<void>((resolve, reject) => {
-      probe.once("error", reject);
-      probe.listen(port, "127.0.0.1", () => resolve());
-    });
-    await new Promise<void>((resolve, reject) =>
-      probe.close((err) => (err ? reject(err) : resolve())),
-    );
-  });
-});

src/gateway/server.models-voicewake-misc.test.ts

@@ -1,19 +1,93 @@
 import fs from "node:fs/promises";
+import { createServer } from "node:net";
 import os from "node:os";
 import path from "node:path";
-import { describe, expect, test } from "vitest";
+import { afterAll, beforeAll, describe, expect, test } from "vitest";
 import { WebSocket } from "ws";
+
+import { getChannelPlugin } from "../channels/plugins/index.js";
+import type { ChannelOutboundAdapter } from "../channels/plugins/types.js";
+import { resolveCanvasHostUrl } from "../infra/canvas-host-url.js";
+import { GatewayLockError } from "../infra/gateway-lock.js";
+import type { PluginRegistry } from "../plugins/registry.js";
+import { getActivePluginRegistry, setActivePluginRegistry } from "../plugins/runtime.js";
+import { createOutboundTestPlugin } from "../test-utils/channel-plugins.js";
+import { GATEWAY_CLIENT_MODES, GATEWAY_CLIENT_NAMES } from "../utils/message-channel.js";
 import {
   connectOk,
+  getFreePort,
   installGatewayTestHooks,
+  occupyPort,
   onceMessage,
   piSdkMock,
   rpcReq,
+  startGatewayServer,
   startServerWithClient,
+  testState,
+  testTailnetIPv4,
 } from "./test-helpers.js";
-import { GATEWAY_CLIENT_MODES, GATEWAY_CLIENT_NAMES } from "../utils/message-channel.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
+
+let server: Awaited<ReturnType<typeof startServerWithClient>>["server"];
+let ws: WebSocket;
+let port: number;
+
+beforeAll(async () => {
+  const started = await startServerWithClient();
+  server = started.server;
+  ws = started.ws;
+  port = started.port;
+  await connectOk(ws);
+});
+
+afterAll(async () => {
+  ws.close();
+  await server.close();
+});
+
+const whatsappOutbound: ChannelOutboundAdapter = {
+  deliveryMode: "direct",
+  sendText: async ({ deps, to, text }) => {
+    if (!deps?.sendWhatsApp) {
+      throw new Error("Missing sendWhatsApp dep");
+    }
+    return { channel: "whatsapp", ...(await deps.sendWhatsApp(to, text, {})) };
+  },
+  sendMedia: async ({ deps, to, text, mediaUrl }) => {
+    if (!deps?.sendWhatsApp) {
+      throw new Error("Missing sendWhatsApp dep");
+    }
+    return { channel: "whatsapp", ...(await deps.sendWhatsApp(to, text, { mediaUrl })) };
+  },
+};
+
+const whatsappPlugin = createOutboundTestPlugin({
+  id: "whatsapp",
+  outbound: whatsappOutbound,
+  label: "WhatsApp",
+});
+
+const createRegistry = (channels: PluginRegistry["channels"]): PluginRegistry => ({
+  plugins: [],
+  tools: [],
+  channels,
+  providers: [],
+  gatewayHandlers: {},
+  httpHandlers: [],
+  cliRegistrars: [],
+  services: [],
+  diagnostics: [],
+});
+
+const whatsappRegistry = createRegistry([
+  {
+    pluginId: "whatsapp",
+    source: "test",
+    plugin: whatsappPlugin,
+  },
+]);
+const emptyRegistry = createRegistry([]);
 
 describe("gateway server models + voicewake", () => {
   const setTempHome = (homeDir: string) => {
@@ -68,9 +142,6 @@ describe("gateway server models + voicewake", () => {
       const homeDir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-home-"));
       const restoreHome = setTempHome(homeDir);
 
-      const { server, ws } = await startServerWithClient();
-      await connectOk(ws);
-
       const initial = await rpcReq<{ triggers: string[] }>(ws, "voicewake.get");
       expect(initial.ok).toBe(true);
       expect(initial.payload?.triggers).toEqual(["clawd", "claude", "computer"]);
@@ -104,9 +175,6 @@ describe("gateway server models + voicewake", () => {
       expect(onDisk.triggers).toEqual(["hi", "there"]);
       expect(typeof onDisk.updatedAtMs).toBe("number");
 
-      ws.close();
-      await server.close();
-
       restoreHome();
     },
   );
@@ -115,9 +183,6 @@ describe("gateway server models + voicewake", () => {
     const homeDir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-home-"));
     const restoreHome = setTempHome(homeDir);
 
-    const { server, ws, port } = await startServerWithClient();
-    await connectOk(ws);
-
     const nodeWs = new WebSocket(`ws://127.0.0.1:${port}`);
     await new Promise<void>((resolve) => nodeWs.once("open", resolve));
     const firstEventP = onceMessage<{ type: "event"; event: string; payload?: unknown }>(
@@ -159,9 +224,6 @@ describe("gateway server models + voicewake", () => {
     ]);
 
     nodeWs.close();
-    ws.close();
-    await server.close();
-
     restoreHome();
   });
 
@@ -189,9 +251,6 @@ describe("gateway server models + voicewake", () => {
       },
     ];
 
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res1 = await rpcReq<{
       models: Array<{
         id: string;
@@ -241,23 +300,132 @@ describe("gateway server models + voicewake", () => {
     ]);
 
     expect(piSdkMock.discoverCalls).toBe(1);
-
-    ws.close();
-    await server.close();
   });
 
   test("models.list rejects unknown params", async () => {
     piSdkMock.enabled = true;
     piSdkMock.models = [{ id: "gpt-test-a", name: "A", provider: "openai" }];
 
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
     const res = await rpcReq(ws, "models.list", { extra: true });
     expect(res.ok).toBe(false);
     expect(res.error?.message ?? "").toMatch(/invalid models\.list params/i);
+  });
+});
 
-    ws.close();
-    await server.close();
+describe("gateway server misc", () => {
+  test("hello-ok advertises the gateway port for canvas host", async () => {
+    const prevToken = process.env.CLAWDBOT_GATEWAY_TOKEN;
+    const prevCanvasPort = process.env.CLAWDBOT_CANVAS_HOST_PORT;
+    process.env.CLAWDBOT_GATEWAY_TOKEN = "secret";
+    testTailnetIPv4.value = "100.64.0.1";
+    testState.gatewayBind = "lan";
+    const canvasPort = await getFreePort();
+    testState.canvasHostPort = canvasPort;
+    process.env.CLAWDBOT_CANVAS_HOST_PORT = String(canvasPort);
+
+    const testPort = await getFreePort();
+    const canvasHostUrl = resolveCanvasHostUrl({
+      canvasPort,
+      requestHost: `100.64.0.1:${testPort}`,
+      localAddress: "127.0.0.1",
+    });
+    expect(canvasHostUrl).toBe(`http://100.64.0.1:${canvasPort}`);
+    if (prevToken === undefined) {
+      delete process.env.CLAWDBOT_GATEWAY_TOKEN;
+    } else {
+      process.env.CLAWDBOT_GATEWAY_TOKEN = prevToken;
+    }
+    if (prevCanvasPort === undefined) {
+      delete process.env.CLAWDBOT_CANVAS_HOST_PORT;
+    } else {
+      process.env.CLAWDBOT_CANVAS_HOST_PORT = prevCanvasPort;
+    }
+  });
+
+  test("send dedupes by idempotencyKey", { timeout: 60_000 }, async () => {
+    const prevRegistry = getActivePluginRegistry() ?? emptyRegistry;
+    try {
+      setActivePluginRegistry(whatsappRegistry);
+      expect(getChannelPlugin("whatsapp")).toBeDefined();
+
+      const idem = "same-key";
+      const res1P = onceMessage(ws, (o) => o.type === "res" && o.id === "a1");
+      const res2P = onceMessage(ws, (o) => o.type === "res" && o.id === "a2");
+      const sendReq = (id: string) =>
+        ws.send(
+          JSON.stringify({
+            type: "req",
+            id,
+            method: "send",
+            params: { to: "+15550000000", message: "hi", idempotencyKey: idem },
+          }),
+        );
+      sendReq("a1");
+      sendReq("a2");
+
+      const res1 = await res1P;
+      const res2 = await res2P;
+      expect(res1.ok).toBe(true);
+      expect(res2.ok).toBe(true);
+      expect(res1.payload).toEqual(res2.payload);
+    } finally {
+      setActivePluginRegistry(prevRegistry);
+    }
+  });
+
+  test("auto-enables configured channel plugins on startup", async () => {
+    const configPath = process.env.CLAWDBOT_CONFIG_PATH;
+    if (!configPath) throw new Error("Missing CLAWDBOT_CONFIG_PATH");
+    await fs.mkdir(path.dirname(configPath), { recursive: true });
+    await fs.writeFile(
+      configPath,
+      JSON.stringify(
+        {
+          channels: {
+            discord: {
+              token: "token-123",
+            },
+          },
+        },
+        null,
+        2,
+      ),
+      "utf-8",
+    );
+
+    const autoPort = await getFreePort();
+    const autoServer = await startGatewayServer(autoPort);
+    await autoServer.close();
+
+    const updated = JSON.parse(await fs.readFile(configPath, "utf-8")) as Record<string, unknown>;
+    const plugins = updated.plugins as Record<string, unknown> | undefined;
+    const entries = plugins?.entries as Record<string, unknown> | undefined;
+    const discord = entries?.discord as Record<string, unknown> | undefined;
+    expect(discord?.enabled).toBe(true);
+    expect((updated.channels as Record<string, unknown> | undefined)?.discord).toMatchObject({
+      token: "token-123",
+    });
+  });
+
+  test("refuses to start when port already bound", async () => {
+    const { server: blocker, port: blockedPort } = await occupyPort();
+    await expect(startGatewayServer(blockedPort)).rejects.toBeInstanceOf(GatewayLockError);
+    await expect(startGatewayServer(blockedPort)).rejects.toThrow(/already listening/i);
+    blocker.close();
+  });
+
+  test("releases port after close", async () => {
+    const releasePort = await getFreePort();
+    const releaseServer = await startGatewayServer(releasePort);
+    await releaseServer.close();
+
+    const probe = createServer();
+    await new Promise<void>((resolve, reject) => {
+      probe.once("error", reject);
+      probe.listen(releasePort, "127.0.0.1", () => resolve());
+    });
+    await new Promise<void>((resolve, reject) =>
+      probe.close((err) => (err ? reject(err) : resolve())),
+    );
   });
 });

src/gateway/server.reload.test.ts

@@ -1,5 +1,12 @@
 import { afterEach, beforeEach, describe, expect, it, vi } from "vitest";
-import { getFreePort, installGatewayTestHooks, startGatewayServer } from "./test-helpers.js";
+import {
+  connectOk,
+  getFreePort,
+  installGatewayTestHooks,
+  rpcReq,
+  startGatewayServer,
+  startServerWithClient,
+} from "./test-helpers.js";
 
 const hoisted = vi.hoisted(() => {
   const cronInstances: Array<{
@@ -158,7 +165,7 @@ vi.mock("./config-reload.js", () => ({
   startGatewayConfigReloader: hoisted.startGatewayConfigReloader,
 }));
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
 
 describe("gateway hot reload", () => {
   let prevSkipChannels: string | undefined;
@@ -298,3 +305,15 @@ describe("gateway hot reload", () => {
     await server.close();
   });
 });
+
+describe("gateway agents", () => {
+  it("lists configured agents via agents.list RPC", async () => {
+    const { server, ws } = await startServerWithClient();
+    await connectOk(ws);
+    const res = await rpcReq<{ agents: Array<{ id: string }> }>(ws, "agents.list", {});
+    expect(res.ok).toBe(true);
+    expect(res.payload?.agents.map((agent) => agent.id)).toContain("main");
+    ws.close();
+    await server.close();
+  });
+});

src/gateway/server.roles-allowlist-update.test.ts

@@ -1,15 +1,48 @@
-import { describe, expect, test } from "vitest";
+import fs from "node:fs/promises";
+import os from "node:os";
+import path from "node:path";
+import { afterAll, beforeAll, describe, expect, test, vi } from "vitest";
+import { WebSocket } from "ws";
 
 import { GATEWAY_CLIENT_MODES, GATEWAY_CLIENT_NAMES } from "../utils/message-channel.js";
+import { GatewayClient } from "./client.js";
+
+vi.mock("../infra/update-runner.js", () => ({
+  runGatewayUpdate: vi.fn(async () => ({
+    status: "ok",
+    mode: "git",
+    root: "/repo",
+    steps: [],
+    durationMs: 12,
+  })),
+}));
+
 import {
   connectOk,
   installGatewayTestHooks,
+  onceMessage,
   rpcReq,
   startServerWithClient,
 } from "./test-helpers.js";
-import { GatewayClient } from "./client.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
+
+let server: Awaited<ReturnType<typeof startServerWithClient>>["server"];
+let ws: WebSocket;
+let port: number;
+
+beforeAll(async () => {
+  const started = await startServerWithClient();
+  server = started.server;
+  ws = started.ws;
+  port = started.port;
+  await connectOk(ws);
+});
+
+afterAll(async () => {
+  ws.close();
+  await server.close();
+});
 
 const sleep = (ms: number) => new Promise((resolve) => setTimeout(resolve, ms));
 
@@ -65,11 +98,99 @@ const connectNodeClient = async (params: {
   return client;
 };
 
+async function waitForSignal(check: () => boolean, timeoutMs = 2000) {
+  const start = Date.now();
+  while (Date.now() - start < timeoutMs) {
+    if (check()) return;
+    await new Promise((resolve) => setTimeout(resolve, 10));
+  }
+  throw new Error("timeout");
+}
+
+describe("gateway role enforcement", () => {
+  test("enforces operator and node permissions", async () => {
+    const nodeWs = new WebSocket(`ws://127.0.0.1:${port}`);
+    await new Promise<void>((resolve) => nodeWs.once("open", resolve));
+
+    try {
+      const eventRes = await rpcReq(ws, "node.event", { event: "test", payload: { ok: true } });
+      expect(eventRes.ok).toBe(false);
+      expect(eventRes.error?.message ?? "").toContain("unauthorized role");
+
+      const invokeRes = await rpcReq(ws, "node.invoke.result", {
+        id: "invoke-1",
+        nodeId: "node-1",
+        ok: true,
+      });
+      expect(invokeRes.ok).toBe(false);
+      expect(invokeRes.error?.message ?? "").toContain("unauthorized role");
+
+      await connectOk(nodeWs, {
+        role: "node",
+        client: {
+          id: GATEWAY_CLIENT_NAMES.NODE_HOST,
+          version: "1.0.0",
+          platform: "ios",
+          mode: GATEWAY_CLIENT_MODES.NODE,
+        },
+        commands: [],
+      });
+
+      const binsRes = await rpcReq<{ bins?: unknown[] }>(nodeWs, "skills.bins", {});
+      expect(binsRes.ok).toBe(true);
+      expect(Array.isArray(binsRes.payload?.bins)).toBe(true);
+
+      const statusRes = await rpcReq(nodeWs, "status", {});
+      expect(statusRes.ok).toBe(false);
+      expect(statusRes.error?.message ?? "").toContain("unauthorized role");
+    } finally {
+      nodeWs.close();
+    }
+  });
+});
+
+describe("gateway update.run", () => {
+  test("writes sentinel and schedules restart", async () => {
+    const sigusr1 = vi.fn();
+    process.on("SIGUSR1", sigusr1);
+
+    try {
+      const id = "req-update";
+      ws.send(
+        JSON.stringify({
+          type: "req",
+          id,
+          method: "update.run",
+          params: {
+            sessionKey: "agent:main:whatsapp:dm:+15555550123",
+            restartDelayMs: 0,
+          },
+        }),
+      );
+      const res = await onceMessage<{ ok: boolean; payload?: unknown }>(
+        ws,
+        (o) => o.type === "res" && o.id === id,
+      );
+      expect(res.ok).toBe(true);
+
+      await waitForSignal(() => sigusr1.mock.calls.length > 0);
+      expect(sigusr1).toHaveBeenCalled();
+
+      const sentinelPath = path.join(os.homedir(), ".clawdbot", "restart-sentinel.json");
+      const raw = await fs.readFile(sentinelPath, "utf-8");
+      const parsed = JSON.parse(raw) as {
+        payload?: { kind?: string; stats?: { mode?: string } };
+      };
+      expect(parsed.payload?.kind).toBe("update");
+      expect(parsed.payload?.stats?.mode).toBe("git");
+    } finally {
+      process.off("SIGUSR1", sigusr1);
+    }
+  });
+});
+
 describe("gateway node command allowlist", () => {
   test("enforces command allowlists across node clients", async () => {
-    const { server, ws, port } = await startServerWithClient();
-    await connectOk(ws);
-
     const waitForConnectedCount = async (count: number) => {
       await expect
         .poll(
@@ -96,8 +217,12 @@ describe("gateway node command allowlist", () => {
       return nodeId;
     };
 
+    let systemClient: GatewayClient | undefined;
+    let emptyClient: GatewayClient | undefined;
+    let allowedClient: GatewayClient | undefined;
+
     try {
-      const systemClient = await connectNodeClient({
+      systemClient = await connectNodeClient({
         port,
         commands: ["system.run"],
         instanceId: "node-system-run",
@@ -115,7 +240,7 @@ describe("gateway node command allowlist", () => {
       systemClient.stop();
       await waitForConnectedCount(0);
 
-      const emptyClient = await connectNodeClient({
+      emptyClient = await connectNodeClient({
         port,
         commands: [],
         instanceId: "node-empty",
@@ -138,7 +263,7 @@ describe("gateway node command allowlist", () => {
         new Promise<{ id?: string; nodeId?: string }>((resolve) => {
           resolveInvoke = resolve;
         });
-      const allowedClient = await connectNodeClient({
+      allowedClient = await connectNodeClient({
         port,
         commands: ["canvas.snapshot"],
         instanceId: "node-allowed",
@@ -187,11 +312,10 @@ describe("gateway node command allowlist", () => {
       });
       const invokeNullRes = await invokeNullResP;
       expect(invokeNullRes.ok).toBe(true);
-
-      allowedClient.stop();
     } finally {
-      ws.close();
-      await server.close();
+      systemClient?.stop();
+      emptyClient?.stop();
+      allowedClient?.stop();
     }
   });
 });

src/gateway/server.roles.test.ts

@@ -1,56 +0,0 @@
-import { describe, expect, test } from "vitest";
-import { WebSocket } from "ws";
-
-import { GATEWAY_CLIENT_MODES, GATEWAY_CLIENT_NAMES } from "../utils/message-channel.js";
-import {
-  connectOk,
-  installGatewayTestHooks,
-  rpcReq,
-  startServerWithClient,
-} from "./test-helpers.js";
-
-installGatewayTestHooks();
-
-describe("gateway role enforcement", () => {
-  test("enforces operator and node permissions", async () => {
-    const { server, ws, port } = await startServerWithClient();
-    await connectOk(ws);
-
-    const eventRes = await rpcReq(ws, "node.event", { event: "test", payload: { ok: true } });
-    expect(eventRes.ok).toBe(false);
-    expect(eventRes.error?.message ?? "").toContain("unauthorized role");
-
-    const invokeRes = await rpcReq(ws, "node.invoke.result", {
-      id: "invoke-1",
-      nodeId: "node-1",
-      ok: true,
-    });
-    expect(invokeRes.ok).toBe(false);
-    expect(invokeRes.error?.message ?? "").toContain("unauthorized role");
-
-    const nodeWs = new WebSocket(`ws://127.0.0.1:${port}`);
-    await new Promise<void>((resolve) => nodeWs.once("open", resolve));
-    await connectOk(nodeWs, {
-      role: "node",
-      client: {
-        id: GATEWAY_CLIENT_NAMES.NODE_HOST,
-        version: "1.0.0",
-        platform: "ios",
-        mode: GATEWAY_CLIENT_MODES.NODE,
-      },
-      commands: [],
-    });
-
-    const binsRes = await rpcReq<{ bins?: unknown[] }>(nodeWs, "skills.bins", {});
-    expect(binsRes.ok).toBe(true);
-    expect(Array.isArray(binsRes.payload?.bins)).toBe(true);
-
-    const statusRes = await rpcReq(nodeWs, "status", {});
-    expect(statusRes.ok).toBe(false);
-    expect(statusRes.error?.message ?? "").toContain("unauthorized role");
-
-    nodeWs.close();
-    ws.close();
-    await server.close();
-  });
-});

src/gateway/server.sessions-send.test.ts

@@ -1,6 +1,6 @@
 import fs from "node:fs/promises";
 import path from "node:path";
-import { afterEach, describe, expect, it, vi } from "vitest";
+import { afterAll, beforeAll, describe, expect, it, vi } from "vitest";
 import { createClawdbotTools } from "../agents/clawdbot-tools.js";
 import { resolveSessionTranscriptPath } from "../config/sessions.js";
 import { emitAgentEvent } from "../infra/agent-events.js";
@@ -11,30 +11,38 @@ import {
   startGatewayServer,
 } from "./test-helpers.js";
 
-installGatewayTestHooks();
+installGatewayTestHooks({ scope: "suite" });
 
-const servers: Array<Awaited<ReturnType<typeof startGatewayServer>>> = [];
+let server: Awaited<ReturnType<typeof startGatewayServer>>;
+let gatewayPort: number;
+let prevGatewayPort: string | undefined;
+let prevGatewayToken: string | undefined;
 
-afterEach(async () => {
-  for (const server of servers) {
-    try {
-      await server.close();
-    } catch {
-      /* ignore */
-    }
+beforeAll(async () => {
+  prevGatewayPort = process.env.CLAWDBOT_GATEWAY_PORT;
+  prevGatewayToken = process.env.CLAWDBOT_GATEWAY_TOKEN;
+  gatewayPort = await getFreePort();
+  process.env.CLAWDBOT_GATEWAY_PORT = String(gatewayPort);
+  process.env.CLAWDBOT_GATEWAY_TOKEN = "test-token";
+  server = await startGatewayServer(gatewayPort);
+});
+
+afterAll(async () => {
+  await server.close();
+  if (prevGatewayPort === undefined) {
+    delete process.env.CLAWDBOT_GATEWAY_PORT;
+  } else {
+    process.env.CLAWDBOT_GATEWAY_PORT = prevGatewayPort;
+  }
+  if (prevGatewayToken === undefined) {
+    delete process.env.CLAWDBOT_GATEWAY_TOKEN;
+  } else {
+    process.env.CLAWDBOT_GATEWAY_TOKEN = prevGatewayToken;
   }
-  servers.length = 0;
-  // Add small delay to ensure port is fully released by OS
-  await new Promise((resolve) => setTimeout(resolve, 50));
 });
 
 describe("sessions_send gateway loopback", () => {
   it("returns reply when lifecycle ends before agent.wait", async () => {
-    const port = await getFreePort();
-    vi.stubEnv("CLAWDBOT_GATEWAY_PORT", String(port));
-    vi.stubEnv("CLAWDBOT_GATEWAY_TOKEN", "test-token");
-
-    const server = await startGatewayServer(port);
     const spy = vi.mocked(agentCommand);
     spy.mockImplementation(async (opts) => {
       const params = opts as {
@@ -78,8 +86,6 @@ describe("sessions_send gateway loopback", () => {
       });
     });
 
-    servers.push(server);
-
     const tool = createClawdbotTools().find((candidate) => candidate.name === "sessions_send");
     if (!tool) throw new Error("missing sessions_send tool");
 
@@ -104,12 +110,6 @@ describe("sessions_send gateway loopback", () => {
 
 describe("sessions_send label lookup", () => {
   it("finds session by label and sends message", { timeout: 60_000 }, async () => {
-    const port = await getFreePort();
-    vi.stubEnv("CLAWDBOT_GATEWAY_PORT", String(port));
-    vi.stubEnv("CLAWDBOT_GATEWAY_TOKEN", "test-token");
-
-    const server = await startGatewayServer(port);
-    servers.push(server);
     const spy = vi.mocked(agentCommand);
     spy.mockImplementation(async (opts) => {
       const params = opts as {
@@ -171,13 +171,6 @@ describe("sessions_send label lookup", () => {
   });
 
   it("returns error when label not found", { timeout: 60_000 }, async () => {
-    const port = await getFreePort();
-    vi.stubEnv("CLAWDBOT_GATEWAY_PORT", String(port));
-    vi.stubEnv("CLAWDBOT_GATEWAY_TOKEN", "test-token");
-
-    const server = await startGatewayServer(port);
-    servers.push(server);
-
     const tool = createClawdbotTools().find((candidate) => candidate.name === "sessions_send");
     if (!tool) throw new Error("missing sessions_send tool");
 
@@ -192,13 +185,6 @@ describe("sessions_send label lookup", () => {
   });
 
   it("returns error when neither sessionKey nor label provided", { timeout: 60_000 }, async () => {
-    const port = await getFreePort();
-    vi.stubEnv("CLAWDBOT_GATEWAY_PORT", String(port));
-    vi.stubEnv("CLAWDBOT_GATEWAY_TOKEN", "test-token");
-
-    const server = await startGatewayServer(port);
-    servers.push(server);
-
     const tool = createClawdbotTools().find((candidate) => candidate.name === "sessions_send");
     if (!tool) throw new Error("missing sessions_send tool");
 

src/gateway/server.sessions.gateway-server-sessions.test.ts

@@ -1,122 +0,0 @@
-import fs from "node:fs/promises";
-import os from "node:os";
-import path from "node:path";
-import { describe, expect, test } from "vitest";
-import {
-  connectOk,
-  installGatewayTestHooks,
-  rpcReq,
-  startServerWithClient,
-  testState,
-  writeSessionStore,
-} from "./test-helpers.js";
-
-installGatewayTestHooks();
-
-describe("gateway server sessions", () => {
-  test("filters sessions by agentId", async () => {
-    const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-sessions-agents-"));
-    testState.sessionConfig = {
-      store: path.join(dir, "{agentId}", "sessions.json"),
-    };
-    testState.agentsConfig = {
-      list: [{ id: "home", default: true }, { id: "work" }],
-    };
-    const homeDir = path.join(dir, "home");
-    const workDir = path.join(dir, "work");
-    await fs.mkdir(homeDir, { recursive: true });
-    await fs.mkdir(workDir, { recursive: true });
-    await writeSessionStore({
-      storePath: path.join(homeDir, "sessions.json"),
-      agentId: "home",
-      entries: {
-        main: {
-          sessionId: "sess-home-main",
-          updatedAt: Date.now(),
-        },
-        "discord:group:dev": {
-          sessionId: "sess-home-group",
-          updatedAt: Date.now() - 1000,
-        },
-      },
-    });
-    await writeSessionStore({
-      storePath: path.join(workDir, "sessions.json"),
-      agentId: "work",
-      entries: {
-        main: {
-          sessionId: "sess-work-main",
-          updatedAt: Date.now(),
-        },
-      },
-    });
-
-    const { ws } = await startServerWithClient();
-    await connectOk(ws);
-
-    const homeSessions = await rpcReq<{
-      sessions: Array<{ key: string }>;
-    }>(ws, "sessions.list", {
-      includeGlobal: false,
-      includeUnknown: false,
-      agentId: "home",
-    });
-    expect(homeSessions.ok).toBe(true);
-    expect(homeSessions.payload?.sessions.map((s) => s.key).sort()).toEqual([
-      "agent:home:discord:group:dev",
-      "agent:home:main",
-    ]);
-
-    const workSessions = await rpcReq<{
-      sessions: Array<{ key: string }>;
-    }>(ws, "sessions.list", {
-      includeGlobal: false,
-      includeUnknown: false,
-      agentId: "work",
-    });
-    expect(workSessions.ok).toBe(true);
-    expect(workSessions.payload?.sessions.map((s) => s.key)).toEqual(["agent:work:main"]);
-  });
-
-  test("resolves and patches main alias to default agent main key", async () => {
-    const dir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-sessions-"));
-    const storePath = path.join(dir, "sessions.json");
-    testState.sessionStorePath = storePath;
-    testState.agentsConfig = { list: [{ id: "ops", default: true }] };
-    testState.sessionConfig = { mainKey: "work" };
-
-    await writeSessionStore({
-      storePath,
-      agentId: "ops",
-      mainKey: "work",
-      entries: {
-        main: {
-          sessionId: "sess-ops-main",
-          updatedAt: Date.now(),
-        },
-      },
-    });
-
-    const { ws } = await startServerWithClient();
-    await connectOk(ws);
-    const resolved = await rpcReq<{ ok: true; key: string }>(ws, "sessions.resolve", {
-      key: "main",
-    });
-    expect(resolved.ok).toBe(true);
-    expect(resolved.payload?.key).toBe("agent:ops:work");
-
-    const patched = await rpcReq<{ ok: true; key: string }>(ws, "sessions.patch", {
-      key: "main",
-      thinkingLevel: "medium",
-    });
-    expect(patched.ok).toBe(true);
-    expect(patched.payload?.key).toBe("agent:ops:work");
-
-    const stored = JSON.parse(await fs.readFile(storePath, "utf-8")) as Record<
-      string,
-      { thinkingLevel?: string }
-    >;
-    expect(stored["agent:ops:work"]?.thinkingLevel).toBe("medium");
-    expect(stored.main).toBeUndefined();
-  });
-});

src/gateway/server.update-run.test.ts

@@ -1,75 +0,0 @@
-import fs from "node:fs/promises";
-import os from "node:os";
-import path from "node:path";
-import { describe, expect, it, vi } from "vitest";
-
-vi.mock("../infra/update-runner.js", () => ({
-  runGatewayUpdate: vi.fn(async () => ({
-    status: "ok",
-    mode: "git",
-    root: "/repo",
-    steps: [],
-    durationMs: 12,
-  })),
-}));
-
-import {
-  connectOk,
-  installGatewayTestHooks,
-  onceMessage,
-  startServerWithClient,
-} from "./test-helpers.js";
-
-installGatewayTestHooks();
-
-async function waitForSignal(check: () => boolean, timeoutMs = 2000) {
-  const start = Date.now();
-  while (Date.now() - start < timeoutMs) {
-    if (check()) return;
-    await new Promise((resolve) => setTimeout(resolve, 10));
-  }
-  throw new Error("timeout");
-}
-
-describe("gateway update.run", () => {
-  it("writes sentinel and schedules restart", async () => {
-    const sigusr1 = vi.fn();
-    process.on("SIGUSR1", sigusr1);
-
-    const { server, ws } = await startServerWithClient();
-    await connectOk(ws);
-
-    const id = "req-update";
-    ws.send(
-      JSON.stringify({
-        type: "req",
-        id,
-        method: "update.run",
-        params: {
-          sessionKey: "agent:main:whatsapp:dm:+15555550123",
-          restartDelayMs: 0,
-        },
-      }),
-    );
-    const res = await onceMessage<{ ok: boolean; payload?: unknown }>(
-      ws,
-      (o) => o.type === "res" && o.id === id,
-    );
-    expect(res.ok).toBe(true);
-
-    await waitForSignal(() => sigusr1.mock.calls.length > 0);
-    expect(sigusr1).toHaveBeenCalled();
-
-    const sentinelPath = path.join(os.homedir(), ".clawdbot", "restart-sentinel.json");
-    const raw = await fs.readFile(sentinelPath, "utf-8");
-    const parsed = JSON.parse(raw) as {
-      payload?: { kind?: string; stats?: { mode?: string } };
-    };
-    expect(parsed.payload?.kind).toBe("update");
-    expect(parsed.payload?.stats?.mode).toBe("git");
-
-    ws.close();
-    await server.close();
-    process.off("SIGUSR1", sigusr1);
-  });
-});

src/memory/manager.async-search.test.ts

@@ -75,7 +75,7 @@ describe("memory search async sync", () => {
 
     const resolved = await Promise.race([
       manager.search("hello").then(() => true),
-      new Promise<boolean>((resolve) => setTimeout(() => resolve(false), 50)),
+      new Promise<boolean>((resolve) => setTimeout(() => resolve(false), 1000)),
     ]);
     expect(resolved).toBe(true);
   });

src/memory/manager.batch.test.ts

@@ -30,6 +30,7 @@ describe("memory indexing with OpenAI batches", () => {
   let workspaceDir: string;
   let indexPath: string;
   let manager: MemoryIndexManager | null = null;
+  let setTimeoutSpy: ReturnType<typeof vi.spyOn>;
 
   beforeEach(async () => {
     embedBatch.mockClear();
@@ -37,6 +38,18 @@ describe("memory indexing with OpenAI batches", () => {
     embedBatch.mockImplementation(async (texts: string[]) =>
       texts.map((_text, index) => [index + 1, 0, 0]),
     );
+    const realSetTimeout = setTimeout;
+    setTimeoutSpy = vi.spyOn(global, "setTimeout").mockImplementation(((
+      handler: TimerHandler,
+      timeout?: number,
+      ...args: unknown[]
+    ) => {
+      const delay = typeof timeout === "number" ? timeout : 0;
+      if (delay > 0 && delay <= 2000) {
+        return realSetTimeout(handler, 0, ...args);
+      }
+      return realSetTimeout(handler, delay, ...args);
+    }) as typeof setTimeout);
     workspaceDir = await fs.mkdtemp(path.join(os.tmpdir(), "clawdbot-mem-batch-"));
     indexPath = path.join(workspaceDir, "index.sqlite");
     await fs.mkdir(path.join(workspaceDir, "memory"));
@@ -44,6 +57,7 @@ describe("memory indexing with OpenAI batches", () => {
 
   afterEach(async () => {
     vi.unstubAllGlobals();
+    setTimeoutSpy.mockRestore();
     if (manager) {
       await manager.close();
       manager = null;

src/memory/manager.embedding-batches.test.ts

@@ -44,7 +44,7 @@ describe("memory embedding batches", () => {
 
   it("splits large files across multiple embedding batches", async () => {
     const line = "a".repeat(200);
-    const content = Array.from({ length: 200 }, () => line).join("\n");
+    const content = Array.from({ length: 50 }, () => line).join("\n");
     await fs.writeFile(path.join(workspaceDir, "memory", "2026-01-03.md"), content);
 
     const cfg = {
@@ -78,7 +78,7 @@ describe("memory embedding batches", () => {
 
   it("keeps small files in a single embedding batch", async () => {
     const line = "b".repeat(120);
-    const content = Array.from({ length: 12 }, () => line).join("\n");
+    const content = Array.from({ length: 4 }, () => line).join("\n");
     await fs.writeFile(path.join(workspaceDir, "memory", "2026-01-04.md"), content);
 
     const cfg = {
@@ -109,7 +109,7 @@ describe("memory embedding batches", () => {
 
   it("reports sync progress totals", async () => {
     const line = "c".repeat(120);
-    const content = Array.from({ length: 20 }, () => line).join("\n");
+    const content = Array.from({ length: 8 }, () => line).join("\n");
     await fs.writeFile(path.join(workspaceDir, "memory", "2026-01-05.md"), content);
 
     const cfg = {
@@ -150,7 +150,7 @@ describe("memory embedding batches", () => {
 
   it("retries embeddings on rate limit errors", async () => {
     const line = "d".repeat(120);
-    const content = Array.from({ length: 12 }, () => line).join("\n");
+    const content = Array.from({ length: 4 }, () => line).join("\n");
     await fs.writeFile(path.join(workspaceDir, "memory", "2026-01-06.md"), content);
 
     let calls = 0;
@@ -162,6 +162,19 @@ describe("memory embedding batches", () => {
       return texts.map(() => [0, 1, 0]);
     });
 
+    const realSetTimeout = setTimeout;
+    const setTimeoutSpy = vi.spyOn(global, "setTimeout").mockImplementation(((
+      handler: TimerHandler,
+      timeout?: number,
+      ...args: unknown[]
+    ) => {
+      const delay = typeof timeout === "number" ? timeout : 0;
+      if (delay > 0 && delay <= 2000) {
+        return realSetTimeout(handler, 0, ...args);
+      }
+      return realSetTimeout(handler, delay, ...args);
+    }) as typeof setTimeout);
+
     const cfg = {
       agents: {
         defaults: {
@@ -183,15 +196,18 @@ describe("memory embedding batches", () => {
     expect(result.manager).not.toBeNull();
     if (!result.manager) throw new Error("manager missing");
     manager = result.manager;
-
-    await manager.sync({ force: true });
+    try {
+      await manager.sync({ force: true });
+    } finally {
+      setTimeoutSpy.mockRestore();
+    }
 
     expect(calls).toBe(3);
   }, 10000);
 
   it("retries embeddings on transient 5xx errors", async () => {
     const line = "e".repeat(120);
-    const content = Array.from({ length: 12 }, () => line).join("\n");
+    const content = Array.from({ length: 4 }, () => line).join("\n");
     await fs.writeFile(path.join(workspaceDir, "memory", "2026-01-08.md"), content);
 
     let calls = 0;
@@ -203,6 +219,19 @@ describe("memory embedding batches", () => {
       return texts.map(() => [0, 1, 0]);
     });
 
+    const realSetTimeout = setTimeout;
+    const setTimeoutSpy = vi.spyOn(global, "setTimeout").mockImplementation(((
+      handler: TimerHandler,
+      timeout?: number,
+      ...args: unknown[]
+    ) => {
+      const delay = typeof timeout === "number" ? timeout : 0;
+      if (delay > 0 && delay <= 2000) {
+        return realSetTimeout(handler, 0, ...args);
+      }
+      return realSetTimeout(handler, delay, ...args);
+    }) as typeof setTimeout);
+
     const cfg = {
       agents: {
         defaults: {
@@ -224,8 +253,11 @@ describe("memory embedding batches", () => {
     expect(result.manager).not.toBeNull();
     if (!result.manager) throw new Error("manager missing");
     manager = result.manager;
-
-    await manager.sync({ force: true });
+    try {
+      await manager.sync({ force: true });
+    } finally {
+      setTimeoutSpy.mockRestore();
+    }
 
     expect(calls).toBe(3);
   }, 10000);

src/plugins/loader.test.ts

@@ -98,8 +98,34 @@ describe("loadClawdbotPlugins", () => {
     expect(enabled?.status).toBe("loaded");
   });
 
-  it("loads bundled telegram plugin when enabled", { timeout: 120_000 }, () => {
-    process.env.CLAWDBOT_BUNDLED_PLUGINS_DIR = path.join(process.cwd(), "extensions");
+  it("loads bundled telegram plugin when enabled", () => {
+    const bundledDir = makeTempDir();
+    writePlugin({
+      id: "telegram",
+      body: `export default { id: "telegram", register(api) {
+  api.registerChannel({
+    plugin: {
+      id: "telegram",
+      meta: {
+        id: "telegram",
+        label: "Telegram",
+        selectionLabel: "Telegram",
+        docsPath: "/channels/telegram",
+        blurb: "telegram channel"
+      },
+      capabilities: { chatTypes: ["direct"] },
+      config: {
+        listAccountIds: () => [],
+        resolveAccount: () => ({ accountId: "default" })
+      },
+      outbound: { deliveryMode: "direct" }
+    }
+  });
+} };`,
+      dir: bundledDir,
+      filename: "telegram.ts",
+    });
+    process.env.CLAWDBOT_BUNDLED_PLUGINS_DIR = bundledDir;
 
     const registry = loadClawdbotPlugins({
       cache: false,

src/slack/actions.ts

@@ -1,8 +1,9 @@
-import { WebClient } from "@slack/web-api";
+import type { WebClient } from "@slack/web-api";
 
 import { loadConfig } from "../config/config.js";
 import { logVerbose } from "../globals.js";
 import { resolveSlackAccount } from "./accounts.js";
+import { createSlackWebClient } from "./client.js";
 import { sendMessageSlack } from "./send.js";
 import { resolveSlackBotToken } from "./token.js";
 
@@ -56,7 +57,7 @@ function normalizeEmoji(raw: string) {
 
 async function getClient(opts: SlackActionClientOpts = {}) {
   const token = resolveToken(opts.token, opts.accountId);
-  return opts.client ?? new WebClient(token);
+  return opts.client ?? createSlackWebClient(token);
 }
 
 async function resolveBotUserId(client: WebClient) {

src/slack/client.test.ts

@@ -0,0 +1,46 @@
+import { describe, expect, it, vi } from "vitest";
+
+vi.mock("@slack/web-api", () => {
+  const WebClient = vi.fn(function WebClientMock(
+    this: Record<string, unknown>,
+    token: string,
+    options?: Record<string, unknown>,
+  ) {
+    this.token = token;
+    this.options = options;
+  });
+  return { WebClient };
+});
+
+const slackWebApi = await import("@slack/web-api");
+const { createSlackWebClient, resolveSlackWebClientOptions, SLACK_DEFAULT_RETRY_OPTIONS } =
+  await import("./client.js");
+
+const WebClient = slackWebApi.WebClient as unknown as ReturnType<typeof vi.fn>;
+
+describe("slack web client config", () => {
+  it("applies the default retry config when none is provided", () => {
+    const options = resolveSlackWebClientOptions();
+
+    expect(options.retryConfig).toEqual(SLACK_DEFAULT_RETRY_OPTIONS);
+  });
+
+  it("respects explicit retry config overrides", () => {
+    const customRetry = { retries: 0 };
+    const options = resolveSlackWebClientOptions({ retryConfig: customRetry });
+
+    expect(options.retryConfig).toBe(customRetry);
+  });
+
+  it("passes merged options into WebClient", () => {
+    createSlackWebClient("xoxb-test", { timeout: 1234 });
+
+    expect(WebClient).toHaveBeenCalledWith(
+      "xoxb-test",
+      expect.objectContaining({
+        timeout: 1234,
+        retryConfig: SLACK_DEFAULT_RETRY_OPTIONS,
+      }),
+    );
+  });
+});

src/slack/client.ts

@@ -0,0 +1,20 @@
+import { type RetryOptions, type WebClientOptions, WebClient } from "@slack/web-api";
+
+export const SLACK_DEFAULT_RETRY_OPTIONS: RetryOptions = {
+  retries: 2,
+  factor: 2,
+  minTimeout: 500,
+  maxTimeout: 3000,
+  randomize: true,
+};
+
+export function resolveSlackWebClientOptions(options: WebClientOptions = {}): WebClientOptions {
+  return {
+    ...options,
+    retryConfig: options.retryConfig ?? SLACK_DEFAULT_RETRY_OPTIONS,
+  };
+}
+
+export function createSlackWebClient(token: string, options: WebClientOptions = {}) {
+  return new WebClient(token, resolveSlackWebClientOptions(options));
+}

src/slack/directory-live.ts

@@ -1,4 +1,4 @@
-import { WebClient } from "@slack/web-api";
+import { createSlackWebClient } from "./client.js";
 
 import type { ChannelDirectoryEntry } from "../channels/plugins/types.js";
 import type { DirectoryConfigParams } from "../channels/plugins/directory-config.js";
@@ -61,7 +61,7 @@ export async function listSlackDirectoryPeersLive(
 ): Promise<ChannelDirectoryEntry[]> {
   const token = resolveReadToken(params);
   if (!token) return [];
-  const client = new WebClient(token);
+  const client = createSlackWebClient(token);
   const query = normalizeQuery(params.query);
   const members: SlackUser[] = [];
   let cursor: string | undefined;
@@ -119,7 +119,7 @@ export async function listSlackDirectoryGroupsLive(
 ): Promise<ChannelDirectoryEntry[]> {
   const token = resolveReadToken(params);
   if (!token) return [];
-  const client = new WebClient(token);
+  const client = createSlackWebClient(token);
   const query = normalizeQuery(params.query);
   const channels: SlackChannel[] = [];
   let cursor: string | undefined;

src/slack/monitor/provider.ts

@@ -17,6 +17,7 @@ import { resolveSlackChannelAllowlist } from "../resolve-channels.js";
 import { resolveSlackUserAllowlist } from "../resolve-users.js";
 import { resolveSlackAppToken, resolveSlackBotToken } from "../token.js";
 import { normalizeSlackWebhookPath, registerSlackHttpHandler } from "../http/index.js";
+import { resolveSlackWebClientOptions } from "../client.js";
 import { resolveSlackSlashCommandConfig } from "./commands.js";
 import { createSlackMonitorContext } from "./context.js";
 import { registerSlackMonitorEvents } from "./events.js";
@@ -130,16 +131,19 @@ export async function monitorSlackProvider(opts: MonitorSlackOpts = {}) {
           endpoints: slackWebhookPath,
         })
       : null;
+  const clientOptions = resolveSlackWebClientOptions();
   const app = new App(
     slackMode === "socket"
       ? {
           token: botToken,
           appToken,
           socketMode: true,
+          clientOptions,
         }
       : {
           token: botToken,
           receiver: receiver ?? undefined,
+          clientOptions,
         },
   );
   const slackHttpHandler =

src/slack/probe.ts

@@ -1,4 +1,4 @@
-import { WebClient } from "@slack/web-api";
+import { createSlackWebClient } from "./client.js";
 
 export type SlackProbe = {
   ok: boolean;
@@ -21,7 +21,7 @@ function withTimeout<T>(promise: Promise<T>, timeoutMs: number): Promise<T> {
 }
 
 export async function probeSlack(token: string, timeoutMs = 2500): Promise<SlackProbe> {
-  const client = new WebClient(token);
+  const client = createSlackWebClient(token);
   const start = Date.now();
   try {
     const result = await withTimeout(client.auth.test(), timeoutMs);

src/slack/resolve-channels.ts

@@ -1,4 +1,6 @@
-import { WebClient } from "@slack/web-api";
+import type { WebClient } from "@slack/web-api";
+
+import { createSlackWebClient } from "./client.js";
 
 export type SlackChannelLookup = {
   id: string;
@@ -84,7 +86,7 @@ export async function resolveSlackChannelAllowlist(params: {
   entries: string[];
   client?: WebClient;
 }): Promise<SlackChannelResolution[]> {
-  const client = params.client ?? new WebClient(params.token);
+  const client = params.client ?? createSlackWebClient(params.token);
   const channels = await listSlackChannels(client);
   const results: SlackChannelResolution[] = [];
 

src/slack/resolve-users.ts

@@ -1,4 +1,6 @@
-import { WebClient } from "@slack/web-api";
+import type { WebClient } from "@slack/web-api";
+
+import { createSlackWebClient } from "./client.js";
 
 export type SlackUserLookup = {
   id: string;
@@ -101,7 +103,7 @@ export async function resolveSlackUserAllowlist(params: {
   entries: string[];
   client?: WebClient;
 }): Promise<SlackUserResolution[]> {
-  const client = params.client ?? new WebClient(params.token);
+  const client = params.client ?? createSlackWebClient(params.token);
   const users = await listSlackUsers(client);
   const results: SlackUserResolution[] = [];
 

src/slack/scopes.ts

@@ -1,4 +1,6 @@
-import { WebClient } from "@slack/web-api";
+import type { WebClient } from "@slack/web-api";
+
+import { createSlackWebClient } from "./client.js";
 
 export type SlackScopesResult = {
   ok: boolean;
@@ -81,7 +83,7 @@ export async function fetchSlackScopes(
   token: string,
   timeoutMs: number,
 ): Promise<SlackScopesResult> {
-  const client = new WebClient(token, { timeout: timeoutMs });
+  const client = createSlackWebClient(token, { timeout: timeoutMs });
   const attempts: SlackScopesSource[] = ["auth.scopes", "apps.permissions.info"];
   const errors: string[] = [];
 

src/slack/send.ts

@@ -1,4 +1,4 @@
-import { type FilesUploadV2Arguments, WebClient } from "@slack/web-api";
+import { type FilesUploadV2Arguments, type WebClient } from "@slack/web-api";
 
 import { resolveTextChunkLimit } from "../auto-reply/chunk.js";
 import { loadConfig } from "../config/config.js";
@@ -6,6 +6,7 @@ import { logVerbose } from "../globals.js";
 import { loadWebMedia } from "../web/media.js";
 import type { SlackTokenSource } from "./accounts.js";
 import { resolveSlackAccount } from "./accounts.js";
+import { createSlackWebClient } from "./client.js";
 import { markdownToSlackMrkdwnChunks } from "./format.js";
 import { parseSlackTarget } from "./targets.js";
 import { resolveSlackBotToken } from "./token.js";
@@ -137,7 +138,7 @@ export async function sendMessageSlack(
     fallbackToken: account.botToken,
     fallbackSource: account.botTokenSource,
   });
-  const client = opts.client ?? new WebClient(token);
+  const client = opts.client ?? createSlackWebClient(token);
   const recipient = parseRecipient(to);
   const { channelId } = await resolveChannelId(client, recipient);
   const textLimit = resolveTextChunkLimit(cfg, "slack", account.accountId);

src/telegram/bot.media.downloads-media-file-path-no-file-download.test.ts

@@ -21,6 +21,7 @@ const apiStub: ApiStub = {
 };
 
 beforeEach(() => {
+  vi.useRealTimers();
   resetInboundDedupe();
 });
 

src/telegram/bot.test.ts

@@ -6,7 +6,12 @@ import {
   listNativeCommandSpecs,
   listNativeCommandSpecsForConfig,
 } from "../auto-reply/commands-registry.js";
-import { listSkillCommandsForAgents } from "../auto-reply/skill-commands.js";
+const { listSkillCommandsForAgents } = vi.hoisted(() => ({
+  listSkillCommandsForAgents: vi.fn(() => []),
+}));
+vi.mock("../auto-reply/skill-commands.js", () => ({
+  listSkillCommandsForAgents,
+}));
 import { resetInboundDedupe } from "../auto-reply/reply/inbound-dedupe.js";
 import * as replyModule from "../auto-reply/reply.js";
 import { expectInboundContextContract } from "../../test/helpers/inbound-contract.js";

src/tui/tui.ts

@@ -452,11 +452,16 @@ export async function runTui(opts: TuiOptions) {
     const reasoning = sessionInfo.reasoningLevel ?? "off";
     const reasoningLabel =
       reasoning === "on" ? "reasoning" : reasoning === "stream" ? "reasoning:stream" : null;
-    footer.setText(
-      theme.dim(
-        `agent ${agentLabel} | session ${sessionLabel} | ${modelLabel} | think ${think} | verbose ${verbose}${reasoningLabel ? ` | ${reasoningLabel}` : ""} | ${tokens}`,
-      ),
-    );
+    const footerParts = [
+      `agent ${agentLabel}`,
+      `session ${sessionLabel}`,
+      modelLabel,
+      think !== "off" ? `think ${think}` : null,
+      verbose !== "off" ? `verbose ${verbose}` : null,
+      reasoningLabel,
+      tokens,
+    ].filter(Boolean);
+    footer.setText(theme.dim(footerParts.join(" | ")));
   };
 
   const { openOverlay, closeOverlay } = createOverlayHandlers(tui, editor);

src/wizard/onboarding.ts

@@ -356,6 +356,10 @@ export async function runOnboardingWizard(
     prompter,
     runtime,
     setDefaultModel: true,
+    opts: {
+      tokenProvider: opts.tokenProvider,
+      token: opts.authChoice === "apiKey" && opts.token ? opts.token : undefined,
+    },
   });
   nextConfig = authResult.config;
 

ui/src/ui/app-tool-stream.ts

@@ -142,7 +142,6 @@ export type CompactionStatus = {
   active: boolean;
   startedAt: number | null;
   completedAt: number | null;
-  retryingAt: number | null;
 };
 
 type CompactionHost = ToolStreamHost & {
@@ -155,7 +154,6 @@ const COMPACTION_TOAST_DURATION_MS = 5000;
 export function handleCompactionEvent(host: CompactionHost, payload: AgentEventPayload) {
   const data = payload.data ?? {};
   const phase = typeof data.phase === "string" ? data.phase : "";
-  const willRetry = Boolean(data.willRetry);
   
   // Clear any existing timer
   if (host.compactionClearTimer != null) {
@@ -168,14 +166,12 @@ export function handleCompactionEvent(host: CompactionHost, payload: AgentEventP
       active: true,
       startedAt: Date.now(),
       completedAt: null,
-      retryingAt: null,
     };
   } else if (phase === "end") {
     host.compactionStatus = {
-      active: willRetry,
+      active: false,
       startedAt: host.compactionStatus?.startedAt ?? null,
-      completedAt: willRetry ? null : Date.now(),
-      retryingAt: willRetry ? Date.now() : null,
+      completedAt: Date.now(),
     };
     // Auto-clear the toast after duration
     host.compactionClearTimer = window.setTimeout(() => {
@@ -187,19 +183,13 @@ export function handleCompactionEvent(host: CompactionHost, payload: AgentEventP
 
 export function handleAgentEvent(host: ToolStreamHost, payload?: AgentEventPayload) {
   if (!payload) return;
-
+  
   // Handle compaction events
   if (payload.stream === "compaction") {
-    const sessionKey =
-      typeof payload.sessionKey === "string" ? payload.sessionKey : undefined;
-    if (sessionKey && sessionKey !== host.sessionKey) return;
-    if (!sessionKey && host.chatRunId && payload.runId !== host.chatRunId) return;
-    if (host.chatRunId && payload.runId !== host.chatRunId) return;
-    if (!host.chatRunId) return;
     handleCompactionEvent(host as CompactionHost, payload);
     return;
   }
-
+  
   if (payload.stream !== "tool") return;
   const sessionKey =
     typeof payload.sessionKey === "string" ? payload.sessionKey : undefined;

ui/src/ui/views/chat.ts

@@ -20,7 +20,6 @@ export type CompactionIndicatorStatus = {
   active: boolean;
   startedAt: number | null;
   completedAt: number | null;
-  retryingAt: number | null;
 };
 
 export type ChatProps = {
@@ -81,17 +80,6 @@ function renderCompactionIndicator(status: CompactionIndicatorStatus | null | un
     `;
   }
   
-  if (status.retryingAt) {
-    const elapsed = Date.now() - status.retryingAt;
-    if (elapsed < COMPACTION_TOAST_DURATION_MS) {
-      return html`
-        <div class="callout info compaction-indicator compaction-indicator--active">
-          🧹 Retrying compaction...
-        </div>
-      `;
-    }
-  }
-
   // Show "compaction complete" briefly after completion
   if (status.completedAt) {
     const elapsed = Date.now() - status.completedAt;