@integrity-labs/agt-cli 0.28.215 → 0.28.217

package/dist/bin/agt.js

@@ -38,7 +38,7 @@ import {
   success,
   table,
   warn
-} from "../chunk-4QCJLG3Q.js";
+} from "../chunk-U6RV5FVC.js";
 import {
   CHANNEL_REGISTRY,
   DEFAULT_FRAMEWORK,
@@ -4826,7 +4826,7 @@ import { execFileSync, execSync } from "child_process";
 import { existsSync as existsSync10, realpathSync as realpathSync2 } from "fs";
 import chalk18 from "chalk";
 import ora16 from "ora";
-var cliVersion = true ? "0.28.215" : "dev";
+var cliVersion = true ? "0.28.217" : "dev";
 async function fetchLatestVersion() {
   const host2 = getHost();
   if (!host2) return null;
@@ -5840,7 +5840,7 @@ function handleError(err) {
 }
 
 // src/bin/agt.ts
-var cliVersion2 = true ? "0.28.215" : "dev";
+var cliVersion2 = true ? "0.28.217" : "dev";
 var program = new Command();
 program.name("agt").description("Augmented CLI \u2014 agent provisioning and management").version(cliVersion2).option("--json", "Emit machine-readable JSON output (suppress spinners and colors)").option("--skip-update-check", "Skip the automatic update check on startup");
 program.hook("preAction", async (thisCommand, actionCommand) => {

package/dist/{chunk-4QCJLG3Q.js → chunk-U6RV5FVC.js}

@@ -5529,7 +5529,7 @@ function requireHost() {
 }
 
 // src/lib/api-client.ts
-var agtCliVersion = true ? "0.28.215" : "dev";
+var agtCliVersion = true ? "0.28.217" : "dev";
 var lastConfigHash = null;
 function setConfigHash(hash) {
   lastConfigHash = hash && hash.length > 0 ? hash : null;
@@ -7581,4 +7581,4 @@ export {
   managerInstallSystemUnitCommand,
   managerUninstallSystemUnitCommand
 };
-//# sourceMappingURL=chunk-4QCJLG3Q.js.map
+//# sourceMappingURL=chunk-U6RV5FVC.js.map

package/dist/lib/manager-worker.js

@@ -36,7 +36,7 @@ import {
   requireHost,
   safeWriteJsonAtomic,
   setConfigHash
-} from "../chunk-4QCJLG3Q.js";
+} from "../chunk-U6RV5FVC.js";
 import {
   getProjectDir as getProjectDir2,
   getReadyTasks,
@@ -685,6 +685,11 @@ var McpFlapDampener = class {
   now;
   history = /* @__PURE__ */ new Map();
   flapping = /* @__PURE__ */ new Set();
+  // ENG-7239: timestamp of the most recent poll on which each (agent, channel)
+  // was flapping. Drives `flappedWithin` so the caller can exempt a managed-MCP
+  // churn restart from the ENG-5441 breaker for a short cool-off after a flap;
+  // see that method for the Kylie/dig (2026-06-29) rationale.
+  lastFlapAt = /* @__PURE__ */ new Map();
   constructor(opts = {}) {
     this.windowMs = opts.windowMs ?? readEnvNumber2("AGT_MCP_FLAP_WINDOW_MS", DEFAULT_WINDOW_MS2);
     this.minTransitions = opts.minTransitions ?? readEnvNumber2("AGT_MCP_FLAP_MIN_TRANSITIONS", DEFAULT_MIN_TRANSITIONS);
@@ -742,9 +747,11 @@ var McpFlapDampener = class {
       this.flapping.add(k);
       onset = true;
     }
+    const nowFlapping = this.flapping.has(k);
+    if (nowFlapping) this.lastFlapAt.set(k, at);
     const reported = tier === "slow" ? slow : fast;
     return {
-      flapping: this.flapping.has(k),
+      flapping: nowFlapping,
       distinctStates: reported.distinctStates,
       transitions: reported.transitions,
       windowCount: prior.length,
@@ -776,6 +783,27 @@ var McpFlapDampener = class {
   isFlapping(codeName, channel) {
     return this.flapping.has(this.key(codeName, channel));
   }
+  /**
+   * ENG-7239: true if (codeName, channel) was flapping within the last
+   * `withinMs`. Lets the caller exempt a managed-MCP-churn session restart from
+   * the ENG-5441 circuit breaker during (and for a short cool-off after) a
+   * flap episode.
+   *
+   * Why a cool-off and not just `isFlapping`: the dampener recovers eagerly, on
+   * the first stable poll (ENG-6512), so during an ACTIVE oscillation a single
+   * transient poll where the desired set momentarily repeats clears `flapping`
+   * for that instant. If the very next poll flips again, that flip's restart
+   * would (pre-fix) be treated as a genuine change and counted toward the
+   * breaker; three such flips in 10 min auto-paused Kylie (org dig) on
+   * 2026-06-29 and left her unreachable on every channel. Keying the breaker
+   * exemption off "flapped within the cool-off" instead of the exact-instant
+   * `isFlapping` closes that race. Mirrors ENG-5547's single-dead-MCP reaper
+   * exemption ("restart, but don't count toward the agent-wide breaker").
+   */
+  flappedWithin(codeName, channel, withinMs) {
+    const at = this.lastFlapAt.get(this.key(codeName, channel));
+    return at !== void 0 && this.now() - at <= withinMs;
+  }
   /** Drop all state for an agent (called on deprovision / agent removal). */
   forget(codeName) {
     const prefix = `${codeName}\0`;
@@ -785,10 +813,29 @@ var McpFlapDampener = class {
     for (const k of [...this.flapping]) {
       if (k.startsWith(prefix)) this.flapping.delete(k);
     }
+    for (const k of [...this.lastFlapAt.keys()]) {
+      if (k.startsWith(prefix)) this.lastFlapAt.delete(k);
+    }
   }
 };
 var FLAP_CHANNEL_INTEGRATIONS = "integrations";
 var FLAP_CHANNEL_MANAGED_MCP = "managed-mcp";
+var DEFAULT_FLAP_BREAKER_COOLDOWN_MS = 6e5;
+var MANAGED_MCP_SERVER_PREFIXES = [
+  "composio_",
+  "one_",
+  "pipedream_",
+  "nango_",
+  "paragon_",
+  "composio-",
+  "one-",
+  "pipedream-",
+  "nango-",
+  "paragon-"
+];
+function isManagedMcpServerKey(key) {
+  return MANAGED_MCP_SERVER_PREFIXES.some((p) => key.startsWith(p));
+}
 
 // src/lib/auto-resume.ts
 var DEFAULT_QUIET_MS = 18e5;
@@ -5365,6 +5412,9 @@ var lastChannelSweepAt = 0;
 var config = null;
 var running = false;
 var pollTimer = null;
+var livenessTimer = null;
+var livenessInFlight = false;
+var LIVENESS_HEARTBEAT_INTERVAL_MS = 3e4;
 var PANE_TAIL_PREVIEW_LINES = 5;
 function extractCharterTelegramPeers(rawContent, gateContext) {
   if (!rawContent || rawContent.length === 0) return [];
@@ -5469,6 +5519,10 @@ var pendingRestartVerifications = /* @__PURE__ */ new Map();
 var restartBreaker = new RestartBreaker();
 var reportedTrips = /* @__PURE__ */ new Map();
 var mcpFlapDampener = new McpFlapDampener();
+var FLAP_BREAKER_COOLDOWN_MS = readEnvNumber2(
+  "AGT_MCP_FLAP_BREAKER_COOLDOWN_MS",
+  DEFAULT_FLAP_BREAKER_COOLDOWN_MS
+);
 function recordConfigChurnEvent(agentId, codeName, channel, signature) {
   api.post("/host/mcp-config-churn", {
     agent_id: agentId,
@@ -5496,6 +5550,12 @@ async function reportReaperEvent(agentId, reaper, action, reason) {
   }
 }
 function recordRestartForBreaker(codeName, reason) {
+  if (reason === "managed-mcp-churn" && mcpFlapDampener.flappedWithin(codeName, FLAP_CHANNEL_MANAGED_MCP, FLAP_BREAKER_COOLDOWN_MS)) {
+    log(
+      `[mcp-flap-dampener] '${codeName}' managed-MCP churn restart NOT counted toward circuit breaker, managed-MCP set flapped within ${Math.round(FLAP_BREAKER_COOLDOWN_MS / 1e3)}s (ENG-7239)`
+    );
+    return;
+  }
   const result = restartBreaker.record(codeName, reason);
   if (!result.tripped || !result.trip) return;
   const trip = result.trip;
@@ -5752,7 +5812,7 @@ function scheduleSessionRestart(codeName, delayMs, reason, breakerReason = "hot-
     runningMcpHashes.delete(codeName);
     recordRestartForBreaker(codeName, breakerReason);
     log(`[hot-reload] Session stopped for '${codeName}' \u2014 will respawn with ${reason}`);
-    if (breakerReason === "hot-reload-mcp" || breakerReason === "bind-remediation" || breakerReason === "channel-set-change") {
+    if (breakerReason === "hot-reload-mcp" || breakerReason === "managed-mcp-churn" || breakerReason === "bind-remediation" || breakerReason === "channel-set-change") {
       const prior = pendingRestartVerifications.get(codeName);
       pendingRestartVerifications.set(codeName, {
         firedAt: Date.now(),
@@ -6064,7 +6124,8 @@ function checkMcpConfigDriftAndScheduleRestart(codeName, projectDir) {
         );
       }
       if (decision.restart && !quarantineOnlyDrift && !flapSuppressed) {
-        scheduleSessionRestart(codeName, 0, ".mcp.json content change (ENG-4897)");
+        const driftBreakerReason = decision.addedOrRemoved.length > 0 && decision.addedOrRemoved.every(isManagedMcpServerKey) ? "managed-mcp-churn" : "hot-reload-mcp";
+        scheduleSessionRestart(codeName, 0, ".mcp.json content change (ENG-4897)", driftBreakerReason);
         runningMcpHashes.delete(codeName);
         runningMcpServerKeys.delete(codeName);
       } else if (flapSuppressed) {
@@ -6163,7 +6224,7 @@ var agentRestartTimezoneInputs = /* @__PURE__ */ new Map();
 var lastVersionCheckAt = 0;
 var VERSION_CHECK_INTERVAL_MS = 5 * 60 * 1e3;
 var lastResponsivenessProbeAt = 0;
-var agtCliVersion = true ? "0.28.215" : "dev";
+var agtCliVersion = true ? "0.28.217" : "dev";
 function resolveBrewPath(execFileSync2) {
   try {
     const out = execFileSync2("which", ["brew"], { timeout: 5e3 }).toString().trim();
@@ -8750,20 +8811,8 @@ async function processAgent(agent, agentStates) {
                 const { readFileSync: readFileSync15 } = await import("fs");
                 const mcpConfig = JSON.parse(readFileSync15(mcpPath, "utf-8"));
                 if (mcpConfig.mcpServers) {
-                  const managedPrefixes = [
-                    "composio_",
-                    "one_",
-                    "pipedream_",
-                    "nango_",
-                    "paragon_",
-                    "composio-",
-                    "one-",
-                    "pipedream-",
-                    "nango-",
-                    "paragon-"
-                  ];
                   for (const key of Object.keys(mcpConfig.mcpServers)) {
-                    if (managedPrefixes.some((p) => key.startsWith(p)) && !expectedServerIds.has(key)) {
+                    if (isManagedMcpServerKey(key) && !expectedServerIds.has(key)) {
                       frameworkAdapter.removeMcpServer(agent.code_name, key);
                       log(`[managed-toolkit] Removed stale MCP server '${key}' for '${agent.code_name}'`);
                     }
@@ -10729,9 +10778,27 @@ async function driveArtifactStreamingInner() {
     if (!liveIds.has(id)) artifactScanners.delete(id);
   }
 }
+async function postLivenessHeartbeat() {
+  if (!running || !config || restartAfterUpgrade) return;
+  if (livenessInFlight) return;
+  livenessInFlight = true;
+  try {
+    const hostId = await getHostId();
+    if (!hostId) return;
+    await api.post("/host/heartbeat", { host_id: hostId, agt_version: agtCliVersion });
+  } catch (err) {
+    log(`[liveness] fast heartbeat failed: ${err.message}`);
+  } finally {
+    livenessInFlight = false;
+  }
+}
 function startPolling() {
   if (!config || running) return;
   running = true;
+  livenessTimer = setInterval(() => {
+    void postLivenessHeartbeat();
+  }, LIVENESS_HEARTBEAT_INTERVAL_MS);
+  livenessTimer.unref();
   void startCaffeinate();
   loadChannelHashCache2();
   hostFlagStore().init();
@@ -10819,6 +10886,10 @@ async function stopPolling(opts = {}) {
     clearTimeout(pollTimer);
     pollTimer = null;
   }
+  if (livenessTimer) {
+    clearInterval(livenessTimer);
+    livenessTimer = null;
+  }
   const shutdownTimer = setTimeout(() => {
     log("Shutdown timeout exceeded (15s), forcing exit");
     process.exit(opts.forcedExitCode ?? 1);