totemlayer 0.1.66 → 0.1.68

package/README.md

@@ -27,7 +27,7 @@ To configure execution later, rerun setup with `--provider` and the adapter's se
 
 ## Connect with automatic listener setup
 
-Configure a local MCP server using command `npx` and arguments `--yes`, `--ignore-scripts`, `--package=https://totemlayer.com/downloads/totemlayer-0.1.66.tgz`, `totem`, `connect`, `--provider`, and `codex` (or `claude` for Claude Code). The host runs the connection itself; the owner does not run a separate listener command. Node.js 22.14 or newer and the selected signed-in client must be available.
+Configure a local MCP server using command `npx` and arguments `--yes`, `--ignore-scripts`, `--package=https://totemlayer.com/downloads/totemlayer-0.1.68.tgz`, `totem`, `connect`, `--provider`, and `codex` (or `claude` for Claude Code). The host runs the connection itself; the owner does not run a separate listener command. Node.js 22.14 or newer and the selected signed-in client must be available.
 
 During browser sign-in, `totem_get_connection_status` reports progress. Normal tools appear after the approved identity is pinned privately. Codex verifies identity, startup and the saved chat callback before receiving work; reconnect preserves bindings and pauses. Missing capabilities leave work queued. A remote HTTP connector cannot install local software.
 
@@ -45,7 +45,7 @@ For a hosted connection, add these settings to your client's MCP configuration:
 For a local stdio connection, install the CLI with Node.js 22.14 or newer:
 
 ```sh
-npm install --global --ignore-scripts https://totemlayer.com/downloads/totemlayer-0.1.66.tgz
+npm install --global --ignore-scripts https://totemlayer.com/downloads/totemlayer-0.1.68.tgz
 ```
 
 Save your agent's connection key in a private file. Protect it with mode 0600 on macOS/Linux or your user's filesystem ACL on Windows. Configure these fields in your client's MCP settings:
@@ -64,7 +64,7 @@ Reload the client if needed, discover the available tools, and call `totem_get_i
 Run the guided installer on the host where your agent will execute work:
 
 ```sh
-npx --yes --package=https://totemlayer.com/downloads/totemlayer-0.1.66.tgz totem install
+npx --yes --package=https://totemlayer.com/downloads/totemlayer-0.1.68.tgz totem install
 ```
 
 The installer opens Totem sign-in and saves the approved connection privately. It checks your configured agent runtime, configures supported clients, and starts the listener when an authorized executor is available. The supervised listener is enabled by default for this setup and starts when you sign in to the computer on hosts with a supported user service manager. It recovers after a process exit and retries temporary connection failures. An explicit stop stays paused across restarts.
@@ -74,7 +74,7 @@ For manual setup, use `totem mcp install` with a supported client and the privat
 For a host without a browser:
 
 ```sh
-npx --yes --package=https://totemlayer.com/downloads/totemlayer-0.1.66.tgz totem login --device-login --no-browser
+npx --yes --package=https://totemlayer.com/downloads/totemlayer-0.1.68.tgz totem login --device-login --no-browser
 ```
 
 Open the printed link on your own device and compare the connection code. Preserve existing MCP entries; inspect conflicts before replacing them. If an executor is unavailable, the saved connection remains available while you complete runtime setup.
@@ -176,3 +176,5 @@ An update changes only the managed code directory. Existing identities, credenti
 `totem update status` shows the installed version and last check. `totem update check` checks now. `totem update pause` disables automatic downloads; `totem update resume` enables them. For a plugin connection, add its `--provider codex` or `--provider claude` selection. `TOTEM_AUTO_UPDATE=0` also disables checks. Pausing updates does not pause message receiving.
 
 Unverified downloads and failed health checks leave working code installed. A receiver that repeatedly crashes immediately after an update falls back to its retained previous version and quarantines the failed version. An offline or sleeping computer checks after it resumes; updates do not wake a closed laptop or add a wake mechanism to an external AI host.
+
+Local Codex setup now defaults to the listener-enabled bound installer with `totem mcp install codex`. The client opens approval and automatically prepares persistent receiving when its MCP connection starts. Verify `totem_get_connection_status` reports `ready=true` and `startupConfigured=true` before reporting installation complete. Use `--transport http` only for an explicit hosted MCP connection; it does not register a host service. Existing settings, identities and owner pauses are preserved.

package/bin/agent-runner.mjs

@@ -1,3 +1,4 @@
+import { processIdentity } from "./runtime-lock.mjs";
 import { startupDiagnostic } from "./startup-diagnostics.mjs";
 import { spawn } from "node:child_process";
 import { randomUUID } from "node:crypto";
@@ -225,11 +226,25 @@ export async function readAgentRuntimeStatus(profile, { isAlive = alive, now = D
   const record = running ?? records[0];
   if (!record) return { running: false, phase: "not_started" };
   const fresh = now() - Date.parse(record.updatedAt) <= 30_000;
+  const receiving =
+    record.listenerHealthRequired !== true ||
+    (typeof record.listenerHeartbeatAt === "string" &&
+      Number.isFinite(Date.parse(record.listenerHeartbeatAt)) &&
+      now() - Date.parse(record.listenerHeartbeatAt) >= -5000 &&
+      now() - Date.parse(record.listenerHeartbeatAt) <= 30_000);
   return {
     running: Boolean(running),
+    listenerReady: Boolean(
+      running &&
+      fresh &&
+      receiving &&
+      record.ready &&
+      !["reconnecting", "starting", "failed", "stopped"].includes(record.phase),
+    ),
     receiveWakeReplyVerified: Boolean(
       running &&
       fresh &&
+      receiving &&
       record.ready === true &&
       !["reconnecting", "starting", "failed", "stopped"].includes(record.phase),
     ),
@@ -238,6 +253,7 @@ export async function readAgentRuntimeStatus(profile, { isAlive = alive, now = D
     ready: Boolean(
       running &&
       fresh &&
+      receiving &&
       record.ready === true &&
       !["blocked", "reconnecting", "starting", "failed", "stopped"].includes(record.phase),
     ),
@@ -251,6 +267,7 @@ export async function readAgentRuntimeStatus(profile, { isAlive = alive, now = D
     runId: record.runId,
     ...(typeof record.packageVersion === "string" ? { packageVersion: record.packageVersion } : {}),
     updatedAt: record.updatedAt,
+    ...(record.listenerHeartbeatAt ? { listenerHeartbeatAt: record.listenerHeartbeatAt } : {}),
     ...(/^[A-Z][A-Z0-9_]{0,79}$/.test(record.errorCode ?? "")
       ? { errorCode: record.errorCode }
       : {}),
@@ -261,8 +278,57 @@ export async function readAgentRuntimeStatus(profile, { isAlive = alive, now = D
   };
 }
 
+/** Recover only this run's verified worker; never signal a reused or unrelated PID. */
+export async function stopStalledAgent(profile, runId, options = {}) {
+  if (!UUID.test(runId)) throw new Error("Invalid stalled run identifier.");
+  const readStatus = options.readStatus ?? readAgentRuntimeStatus;
+  const identity = options.processIdentity ?? processIdentity;
+  const signal = options.processSignal ?? process.kill;
+  const current = await readStatus(profile);
+  if (!current.running || current.runId !== runId || current.phase !== "unresponsive") return false;
+  const directory = await runtimeDirectory(profile);
+  const record = await readPrivateJson(join(directory, `${runId}.json`));
+  const command = `${CLI_PATH} start --profile ${profile.profileName} --run-id ${runId} --json`;
+  const owned = async () => {
+    const actual = await identity(record.pid);
+    if (!actual) return false;
+    if (
+      record.pid === process.pid ||
+      record.runId !== runId ||
+      record.agentId !== profile.agentId ||
+      record.baseUrl !== profile.baseUrl ||
+      typeof record.processIdentity !== "string" ||
+      actual !== record.processIdentity ||
+      !actual.replaceAll('"', "").includes(command)
+    )
+      throw Object.assign(new Error("The stalled worker's ownership could not be verified."), {
+        code: "LISTENER_PROCESS_UNVERIFIED",
+      });
+    return true;
+  };
+  if (!(await owned())) return true;
+  await requestStop(profile, runId);
+  if (!(await owned())) return true;
+  signal(record.pid, "SIGTERM");
+  if (process.platform !== "win32" && (await owned())) signal(record.pid, "SIGCONT");
+  const deadline = Date.now() + (options.stopTimeoutMs ?? 3000);
+  while (Date.now() < deadline) {
+    const state = await readStatus(profile);
+    if (!state.running || state.runId !== runId || !(await owned())) return true;
+    await new Promise((resolve) => setTimeout(resolve, 100));
+  }
+  const state = await readStatus(profile);
+  if (!state.running || state.runId !== runId || !(await owned())) return true;
+  // A fresh heartbeat means the worker recovered; leave it and its journal intact.
+  if (state.phase !== "unresponsive") return false;
+  if (await owned()) signal(record.pid, "SIGKILL");
+  return true;
+}
+
 function eventState(event) {
-  if (event?.type === "runtime_ready") return { ready: true };
+  if (event?.type === "runtime_ready")
+    return { ready: true, listenerHeartbeatAt: new Date().toISOString() };
+  if (event?.type === "runtime_heartbeat") return { listenerHeartbeatAt: new Date().toISOString() };
   const candidate = event?.state ?? event?.status ?? event?.type;
   const phase =
     typeof candidate === "string" ? candidate.replace(/^mission_agent_|^agent_/, "") : "";
@@ -328,6 +394,8 @@ export async function runAgentSession(profile, options = {}) {
     baseUrl: profile.baseUrl,
     startedAt: new Date().toISOString(),
     updatedAt: new Date().toISOString(),
+    processIdentity: await (options.processIdentity ?? processIdentity)(process.pid),
+    listenerHealthRequired: true,
     phase: "starting",
   };
   let writes = Promise.resolve();
@@ -383,7 +451,25 @@ export async function runAgentSession(profile, options = {}) {
   }
   let polling;
   let stopReadFailed = false;
+  let listenerExpired = false;
+  const listenerWatchdogMs = options.listenerWatchdogMs ?? 30_000;
+  if (
+    !Number.isInteger(listenerWatchdogMs) ||
+    listenerWatchdogMs < 50 ||
+    listenerWatchdogMs > 120_000
+  )
+    throw new Error("Invalid listener watchdog deadline.");
   const poll = async () => {
+    if (status.ready && Date.now() - Date.parse(status.listenerHeartbeatAt) > listenerWatchdogMs) {
+      listenerExpired = true;
+      status = {
+        ...status,
+        ready: false,
+        phase: "reconnecting",
+        errorCode: "LISTENER_HEARTBEAT_TIMEOUT",
+      };
+      controller.abort();
+    }
     try {
       const request = await readPrivateJson(join(directory, `${runId}.stop`));
       if (
@@ -463,6 +549,10 @@ export async function runAgentSession(profile, options = {}) {
         },
       });
     }
+    if (listenerExpired)
+      throw Object.assign(new Error("The listener stopped renewing its receive heartbeat."), {
+        code: "LISTENER_HEARTBEAT_TIMEOUT",
+      });
     if (persistenceFailed || stopReadFailed)
       throw new Error("Agent status could not be saved; the worker was stopped.");
   } catch (error) {

package/bin/agent-setup.mjs

@@ -551,7 +551,13 @@ export async function agentSetupMain(args, options = {}) {
       ? `Connection saved. Automatic receiving is not configured. To add an executor, run totem setup --profile ${profile.profileName} --provider <provider> --replace with its settings.`
       : `Run totem status --profile ${profile.profileName} to check your connection and automatic listener.`;
     const report = {
-      ok: (!mcp || mcp.state !== "needs_attention") && automatic?.state !== "needs_attention",
+      ok:
+        (!mcp || mcp.state !== "needs_attention") &&
+        (connectionOnly || flags["no-start"] || io.deferActivation
+          ? true
+          : automatic?.ready === true &&
+            automatic?.persistent === true &&
+            automatic?.running === true),
       profile: profile.profileName,
       agent: identity.address,
       origin: baseUrl,

package/bin/automatic-runtime.mjs

@@ -9,12 +9,13 @@ import { setTimeout as sleep } from "node:timers/promises";
 import { fileURLToPath } from "node:url";
 import { Totem } from "./sdk/index.js";
 import { readAgentCredential, readAgentProfile } from "./agent-profile.mjs";
-import { agentStartMain, readAgentRuntimeStatus, requestStop } from "./agent-runner.mjs";
 import {
-  privateRead,
-  privateWrite,
-  runtimeRoot,
-} from "./runtime-installation.mjs";
+  agentStartMain,
+  readAgentRuntimeStatus,
+  requestStop,
+  stopStalledAgent,
+} from "./agent-runner.mjs";
+import { privateRead, privateWrite, runtimeRoot } from "./runtime-installation.mjs";
 import { resumeDeviceLogin } from "./device-login.mjs";
 
 const PACKAGE_VERSION = JSON.parse(
@@ -93,7 +94,13 @@ export async function superviseOnce(env = process.env, options = {}) {
         await reportActivations(activationClient, activationIds, "paused");
         continue;
       }
-      const state = await (options.readStatus ?? readAgentRuntimeStatus)(profile);
+      let state = await (options.readStatus ?? readAgentRuntimeStatus)(profile);
+      if (state.running && state.phase === "unresponsive") {
+        await (options.recoverStalled ?? stopStalledAgent)(profile, state.runId);
+        state = await (options.readStatus ?? readAgentRuntimeStatus)(profile);
+        if (state.running) continue;
+        if (!(await automaticEnabled(profile))) continue;
+      }
       if (state.running && state.packageVersion === (options.packageVersion ?? PACKAGE_VERSION))
         await reportActivations(
           activationClient,
@@ -150,13 +157,18 @@ export async function superviseOnce(env = process.env, options = {}) {
         throw Object.assign(new Error("START_FAILED"), failure);
       }
       const started = await (options.readStatus ?? readAgentRuntimeStatus)(profile);
+      if (!started.running)
+        throw Object.assign(
+          new Error("The listener process did not remain running after startup."),
+          { code: "LISTENER_START_FAILED" },
+        );
       await reportActivations(
         activationClient,
         activationIds,
         started.ready ? "ready" : "starting",
       );
       await privateWrite(join(profile.stateDirectory, "automatic-health.json"), {
-        state: "running",
+        state: started.ready ? "running" : "starting",
         activationIds,
         attempts: 0,
         retryAt: null,

package/bin/connected-agent-service.mjs

@@ -76,7 +76,7 @@ export async function runConnectedAgentService(settings, dependencies = {}) {
             replyTimeoutMs: 90000,
             pollMs: settings.pollMs ?? DEFAULT_LISTENER_POLL_MS,
             emit(event) {
-              if (event.type === "runtime_ready") emit(event);
+              if (["runtime_ready", "runtime_heartbeat"].includes(event.type)) emit(event);
               else if (event.type === "runtime_reconnecting")
                 report("messages", { type: "reconnecting" });
               else if (event.type === "runtime_resumed") report("messages", { type: "waiting" });

package/bin/installation.mjs

@@ -31,13 +31,13 @@ export async function activateConnection(profile, env = process.env, options = {
       ...(profile.agentTokenEnv ? { TOTEM_AGENT_TOKEN_ENV: profile.agentTokenEnv } : {}),
     });
     if (!enabled) return { state: "paused", ...service, entry: runtime.cli };
-    if (profile.interactiveRequired && !service.persistent)
+    if (!service.persistent)
       return {
         state: "needs_attention",
         ready: false,
         ...service,
         entry: runtime.cli,
-        code: "HARNESS_SETUP_REQUIRED",
+        code: "LISTENER_STARTUP_UNAVAILABLE",
         message:
           "Automatic startup is not registered. Incoming work remains queued until host setup succeeds.",
       };
@@ -89,8 +89,9 @@ export async function activateConnection(profile, env = process.env, options = {
     }
     const status = await (options.readStatus ?? readAgentRuntimeStatus)(profile);
     return {
-      state: status.ready ? "ready" : "needs_attention",
-      ready: Boolean(status.ready),
+      state: status.running && status.ready ? "ready" : "needs_attention",
+      ready: Boolean(status.running && status.ready),
+      running: status.running === true,
       ...service,
       entry: runtime.cli,
     };
@@ -232,14 +233,15 @@ export async function installationMain(args = [], options = {}) {
     return (options.activateConnection ?? activateConnection)(profile, environment, options);
   };
   const reportActivation = (status) => {
+    const ready = status.ready === true && status.persistent === true && status.running === true;
     stdout.write(
       flags.json
         ? JSON.stringify(status) + "\n"
-        : status.ready
+        : ready
           ? "Ready. The listener passed its controlled receive, wake and reply check. This host must remain awake and online.\n"
           : `Runtime ${status.state}. Run totem status for diagnostics; an installed or paused listener is not ready.\n`,
     );
-    return status.ready ? 0 : 1;
+    return ready ? 0 : 1;
   };
   if (environment.TOTEM_CONNECTION_ID) {
     for (const file of ["connection-ready.json", "connection-request.json"]) {
@@ -328,8 +330,10 @@ export async function installationMain(args = [], options = {}) {
     for (const name of saved) results.push({ profile: name, ...(await activate(name)) });
     return reportActivation({
       installed: true,
-      ready: results.every((r) => r.ready),
-      state: results.every((r) => r.ready) ? "ready" : "needs_attention",
+      ready: results.every((r) => r.ready && r.persistent && r.running),
+      persistent: results.every((r) => r.persistent),
+      running: results.every((r) => r.running),
+      state: results.every((r) => r.ready && r.persistent && r.running) ? "ready" : "needs_attention",
       profiles: results,
       service,
     });

package/bin/mcp-install.mjs

@@ -7,7 +7,7 @@ import { resolve, join } from "node:path";
 import { fileURLToPath } from "node:url";
 
 import { installedClientInvocation, localClientEnvironment } from "./platform-client.mjs";
-import { runtimeRoot } from "./runtime-installation.mjs";
+import { runtimeRoot, ensureInstalledRuntime } from "./runtime-installation.mjs";
 import { loadMcpIdentityBinding } from "./mcp-identity-binding.mjs";
 
 const DEFAULT_ORIGIN = "https://totemlayer.com";
@@ -18,7 +18,8 @@ export const MCP_INSTALL_HELP = `Connect Totem to an AI app
 
 Usage:
   totem mcp install claude             Add hosted MCP and open one OAuth approval
-  totem mcp install codex              Add hosted MCP and open one OAuth approval
+  totem mcp install codex              Add local MCP with automatic listener setup
+  totem mcp install codex --transport http  Explicit hosted MCP without a host service
   totem mcp install generic            HTTP settings for other hosts
   totem mcp config                     HTTP settings for any MCP client
   totem mcp config --transport stdio --key-file /absolute/path/to/agent.key
@@ -31,13 +32,18 @@ Options:
   --base-url <origin>  Totem origin (default: https://totemlayer.com)
   --name <name>        MCP name (default: totem, or totem-local with --key-file)
   --key-file <path>    Use an existing private agent key for local stdio instead
+  --transport <type>  http or stdio (Codex defaults to listener-enabled stdio)
   --agent-id <UUID>    Require this credential-derived agent identity
   --environment <env> Require PRODUCTION or SANDBOX
   --codex-thread <ID>   Bind a new Codex listener to this owner-selected chat
   --no-start           Configure tools only; do not start a local listener
   --help               Show this guide
 
-Without --key-file, this uses the client's native hosted MCP and OAuth support.
+Without --key-file, Codex uses the bound browser installer and supervised listener.
+The selected Codex host opens approval and finishes listener setup on connection.
+Settings alone are not setup completion: check totem_get_connection_status until
+ready and startupConfigured are true. Explicit stops remain paused.
+Claude and generic use hosted MCP; --transport http explicitly selects that flow.
 The client saves credentials privately. No Totem website guide, copied key,
 local worker, or model change is required. Existing connections are preserved.
 The host may require reloading tools after adding a server. Call
@@ -87,6 +93,7 @@ function parseInstallArgs(args, environment) {
         "--agent-id",
         "--environment",
         "--codex-thread",
+        "--transport",
       ].includes(option) ||
       !value ||
       value.startsWith("--")
@@ -96,6 +103,13 @@ function parseInstallArgs(args, environment) {
     values[option] = value;
   }
   const keyFile = values["--key-file"] ?? environment.TOTEM_NETWORK_API_KEY_FILE;
+  const transport = values["--transport"] ?? (keyFile || client === "codex" ? "stdio" : "http");
+  if (!["http", "stdio"].includes(transport)) fail("Choose --transport http or stdio.");
+  if (keyFile && transport === "http") fail("Use a private key file only with stdio transport.");
+  if (!keyFile && transport === "stdio" && client !== "codex")
+    fail(
+      "Keyless local listener setup currently requires Codex. Use this host's supported incoming integration or an explicit private key file.",
+    );
   const name = values["--name"] ?? (keyFile ? DEFAULT_NAME : "totem");
   if (!/^[A-Za-z][A-Za-z0-9_-]{0,63}$/.test(name))
     fail(
@@ -128,6 +142,7 @@ function parseInstallArgs(args, environment) {
     fail("--codex-thread requires codex and an owner-selected chat UUID.");
   return {
     client,
+    transport,
     codexThreadId: values["--codex-thread"],
     configDirectory: environment.TOTEM_CONFIG_DIR ? runtimeRoot(environment) : undefined,
     noStart: values["--no-start"] === true || environment.TOTEM_AUTO_START === "0",
@@ -305,7 +320,10 @@ export async function installTotemMcp(args, environment = process.env, runtime =
   const options = parseInstallArgs(args, environment);
   if (options.help) return MCP_INSTALL_HELP;
   runtime = { platform: process.platform, spawnProcess: spawnSync, ...runtime };
-  if (!options.keyFile) return installHostedTotemMcp(options, environment, runtime);
+  if (!options.keyFile)
+    return options.transport === "stdio"
+      ? installBoundTotemMcp(options, environment, runtime)
+      : installHostedTotemMcp(options, environment, runtime);
   await validateKeyFile(options.keyFile, runtime.platform);
 
   if (options.client === "generic")
@@ -440,6 +458,74 @@ export async function installTotemMcp(args, environment = process.env, runtime =
   return `Connected ${options.name} to ${command} at ${options.baseUrl}.${listener} Restart ${command}, then ask it to call totem_get_identity. The Totem CLI package must remain installed.\n`;
 }
 
+/** Persist the self-configuring entry, rather than a tools-only hosted URL.
+ * Authorization and readiness happen inside that exact client connection. */
+async function installBoundTotemMcp(options, environment, runtime) {
+  if (options.codexThreadId)
+    fail("Use the existing private-key setup to bind an explicitly selected Codex chat.");
+  const installed = await (runtime.ensureRuntime ?? ensureInstalledRuntime)(environment);
+  const args = [
+    installed.cli,
+    "connect",
+    "--provider",
+    options.client,
+    "--base-url",
+    options.baseUrl,
+    ...(options.expectedAgentId ? ["--agent-id", options.expectedAgentId] : []),
+    ...(options.expectedEnvironment ? ["--environment", options.expectedEnvironment] : []),
+  ];
+  const savedEnv = options.noStart ? { TOTEM_AUTO_START: "0" } : {};
+  const env = localClientEnvironment(environment);
+  const get = () => runClient(options.client, ["mcp", "get", options.name, "--json"], env, runtime);
+  const matches = (entry) => {
+    try {
+      const value = JSON.parse(entry.stdout),
+        transport = value.transport;
+      return (
+        entry.status === 0 &&
+        value.enabled !== false &&
+        transport?.type === "stdio" &&
+        transport.command === process.execPath &&
+        JSON.stringify(transport.args) === JSON.stringify(args) &&
+        JSON.stringify(Object.entries(transport.env ?? {}).sort()) ===
+          JSON.stringify(Object.entries(savedEnv).sort()) &&
+        (!value.disabled_tools || value.disabled_tools.length === 0) &&
+        (!value.enabled_tools || value.enabled_tools.includes("totem_get_connection_status"))
+      );
+    } catch {
+      return false;
+    }
+  };
+  const existing = get();
+  if (existing.status === 0 && !matches(existing))
+    fail(
+      "This MCP entry already has different or disabled settings. It was preserved. Inspect that exact connection before repairing its automatic receiver; adding tools alone does not complete listener setup.",
+    );
+  if (existing.status !== 0) {
+    if (!/no mcp server|not found|does not exist/i.test(`${existing.stderr}\n${existing.stdout}`))
+      fail("The client could not check its existing MCP settings. No configuration was changed.");
+    const add = runClient(
+      options.client,
+      [
+        "mcp",
+        "add",
+        options.name,
+        ...Object.entries(savedEnv).flatMap(([name, value]) => ["--env", `${name}=${value}`]),
+        "--",
+        process.execPath,
+        ...args,
+      ],
+      env,
+      runtime,
+    );
+    if (add.status !== 0 || !matches(get()))
+      fail(
+        "The listener-enabled MCP settings could not be verified. Inspect this exact entry before retrying.",
+      );
+  }
+  return "Totem's listener-enabled MCP settings are saved. Reload the selected client's MCP tools to open its bound browser approval and configure the persistent listener automatically. Setup is incomplete until totem_get_connection_status reports ready=true and startupConfigured=true; then verify a permitted incoming message and reply. Existing pauses and permissions are preserved.\n";
+}
+
 /** Guided setup owns a distinct entry per profile; the manual installer retains its default. */
 export function managedMcpName(profileName) {
   if (!/^[a-z0-9]+(?:-[a-z0-9]+)*$/.test(profileName) || profileName.length > 64)

package/bin/native/darwin-lifetime.json

@@ -1 +1 @@
-{"sourceSha256":"e15dc6cfd6f828dbcb829fc6fa3f5f6ef51062f45462ec6532b8c52f532c7951","binarySha256":"fa6dad0474b12a9e8d4ba11ec1032166939ea330817a13de6cbdece265edbadf","architectures":["arm64","x86_64"],"minimumMacOS":"13.0"}
+{"sourceSha256":"e15dc6cfd6f828dbcb829fc6fa3f5f6ef51062f45462ec6532b8c52f532c7951","binarySha256":"841e12dec0c0269584f5a13793f9a66471af09db73783fcca3d5e1750582aafc","architectures":["arm64","x86_64"],"minimumMacOS":"13.0"}

package/bin/network-runtime.mjs

@@ -793,6 +793,7 @@ export async function runRuntime(settings) {
         },
         { signal },
       );
+      emit({ type: "runtime_heartbeat" });
       diagnosticsSupported = result?.diagnosticsSupported === true;
       operationalStatesSupported = result?.operationalStatesSupported === true;
       return result;

package/bin/runtime-diagnostics.mjs

@@ -7,6 +7,9 @@ const MISSION_ERROR_CODES = new Map([
 ]);
 
 export const RUNTIME_ERROR_CODES = new Set([
+  "LISTENER_STARTUP_UNAVAILABLE",
+  "LISTENER_HEARTBEAT_TIMEOUT",
+  "LISTENER_PROCESS_UNVERIFIED",
   ...MISSION_ERROR_CODES.values(),
   "AGENT_START_FAILED",
   "AGENT_EXIT_FAILED",

package/bin/runtime-installation.mjs

@@ -562,6 +562,22 @@ export async function installRuntimeService(runtime, env = process.env, options
       await run("systemctl", ["--user", "daemon-reload"]);
       await run("systemctl", ["--user", "enable", "--now", unit]);
       if (runtime.upgraded) await run("systemctl", ["--user", "restart", unit]);
+      const enabled = await run("systemctl", ["--user", "is-enabled", unit]);
+      const active = await run("systemctl", ["--user", "is-active", unit]);
+      const loaded = await run("systemctl", [
+        "--user",
+        "show",
+        unit,
+        "--property=ExecStart",
+        "--value",
+      ]);
+      if (
+        enabled.stdout.trim() !== "enabled" ||
+        active.stdout.trim() !== "active" ||
+        !loaded.stdout.includes(node) ||
+        !loaded.stdout.includes(runtime.supervisor)
+      )
+        throw new Error("The exact persistent listener service is not enabled and active.");
       return acknowledged({ mode: "systemd-user", persistent: true, label: unit, path });
     }
     if (platform === "win32") {
@@ -572,6 +588,17 @@ export async function installRuntimeService(runtime, env = process.env, options
   } catch {
     /* Ephemeral/headless hosts often have no user service manager. */
   }
+  if (windowsRegistration?.mode !== "startup-shortcut")
+    return {
+      mode: "unavailable",
+      persistent: false,
+      code:
+        startupCode ??
+        windowsRegistration?.code ??
+        (platform === "darwin" || platform === "win32"
+          ? "STARTUP_REGISTRATION_FAILED"
+          : "LISTENER_STARTUP_UNAVAILABLE"),
+    };
   const child = (options.spawn ?? spawn)(node, [runtime.supervisor], {
     env: { ...env, TOTEM_CONFIG_DIR: runtime.root },
     stdio: "ignore",
@@ -580,13 +607,5 @@ export async function installRuntimeService(runtime, env = process.env, options
   });
   child.on("error", () => {});
   child.unref();
-  return windowsRegistration?.mode === "startup-shortcut"
-    ? acknowledged({ mode: "startup-shortcut", persistent: true, label: task })
-    : {
-        mode: "background-supervisor",
-        persistent: false,
-        ...(["win32", "darwin"].includes(platform)
-          ? { code: startupCode ?? windowsRegistration?.code ?? "STARTUP_REGISTRATION_FAILED" }
-          : {}),
-      };
+  return acknowledged({ mode: "startup-shortcut", persistent: true, label: task });
 }

package/bin/startup-diagnostics.mjs

@@ -4,6 +4,19 @@ export const HOST_APPROVAL_RECOVERY =
   "Stop retrying. Ask the owner to inspect this command in the host app's approval or automatic review controls. If no approval path is available, contact the host's support with the denial. Resume once after the owner resolves it; do not bypass the policy or change agents.";
 
 const causes = {
+  LISTENER_STARTUP_UNAVAILABLE: [
+    "The host has no verified enabled persistent listener service.",
+    "Restore the host's supported startup service, then rerun setup. A temporary background process is not a complete installation.",
+  ],
+  LISTENER_HEARTBEAT_TIMEOUT: [
+    "The listener stopped renewing its receive heartbeat.",
+    "The supervisor will restart this same saved profile. Inspect network availability if receiving repeatedly stalls.",
+  ],
+  LISTENER_PROCESS_UNVERIFIED: [
+    "The stalled listener process could not be safely identified.",
+    "Inspect this profile's saved process and stop the exact worker before resuming; no unrelated process was signaled.",
+    false,
+  ],
   HOST_APPROVAL_DENIED: [
     "The host denied approval to execute the startup command.",
     HOST_APPROVAL_RECOVERY,

package/bin/totem.mjs

@@ -240,9 +240,9 @@ function help() {
   writeBrand();
   writeHelpBody(`Start here
   Updates               totem update status | check | pause | resume
-  Install once          npm install --global https://totemlayer.com/downloads/totemlayer-0.1.66.tgz
+  Install once          npm install --global https://totemlayer.com/downloads/totemlayer-${CLI_VERSION}.tgz
   Confirm installation totem --version
-  Run one-off           npx --yes --package https://totemlayer.com/downloads/totemlayer-0.1.66.tgz totem --help
+  Run one-off           npx --yes --package https://totemlayer.com/downloads/totemlayer-${CLI_VERSION}.tgz totem --help
 
 Connect once and receive messages and approved Mission work
   Select Codex, Claude, a local model, or an owner-configured CLI/HTTP agent.

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "totemlayer",
-  "version": "0.1.66",
+  "version": "0.1.68",
   "description": "Cross-platform CLI for TotemLayer agent identity and authorization",
   "type": "module",
   "bin": {