agentmetrics-openclaw 0.2.3 → 0.2.4

package/README.md

@@ -1,11 +1,9 @@
 # agentmetrics-openclaw
 
-[![npm](https://img.shields.io/npm/v/agentmetrics-openclaw?color=009E80&label=npm&logo=npm&logoColor=white)](https://www.npmjs.com/package/agentmetrics-openclaw)
-[![License: MIT](https://img.shields.io/badge/license-MIT-009E80)](../LICENSE)
+[![npm](https://img.shields.io/npm/v/agentmetrics-openclaw?color=6366f1&label=npm&logo=npm&logoColor=white)](https://www.npmjs.com/package/agentmetrics-openclaw)
+[![License: MIT](https://img.shields.io/badge/license-MIT-6366f1)](../LICENSE)
 
-Zero-code observability for [OpenClaw](https://openclaw.dev) agents. Install the plugin, set your API key, and every agent session is tracked automatically — tokens, cost, tools, subagents, context health, and reliability.
-
-→ **[agentmetrics.dev/docs/integrations/openclaw](https://agentmetrics.dev/docs/integrations/openclaw)**
+AgentMetrics integration for [OpenClaw](https://openclaw.dev). Install the plugin, point it at your server, and every agent session reports back to your dashboard automatically showing tokens, cost, tools, subagents, context health, and reliability, all without changing your agent code.
 
 ---
 
@@ -13,7 +11,7 @@ Zero-code observability for [OpenClaw](https://openclaw.dev) agents. Install the
 
 - OpenClaw 2026.3.2 or later
 - Node.js 22 or later
-- An AgentMetrics API key — [get one at agentmetrics.dev](https://agentmetrics.dev/signup)
+- A running AgentMetrics server (see the [main README](../../../README.md) for setup)
 
 ---
 
@@ -27,35 +25,60 @@ openclaw plugins install agentmetrics-openclaw
 
 ## Setup
 
-**1. Set your API key**
+**1. Start AgentMetrics** (if not already running)
+
+```bash
+# Docker
+docker compose up
+
+# Or Python CLI
+pip install agentmetrics
+agentmetrics dashboard
+```
+
+**2. Set the server URL** (if not running on localhost)
 
 ```bash
-# macOS / Linux
-echo 'export AGENTMETRICS_API_KEY=am_live_...' >> ~/.bashrc && source ~/.bashrc
+# macOS / Linux, permanent
+echo 'export AGENTMETRICS_BASE_URL=http://your-server:8099' >> ~/.bashrc && source ~/.bashrc
 
 # Windows (PowerShell)
-$Env:AGENTMETRICS_API_KEY = "am_live_..."
+$Env:AGENTMETRICS_BASE_URL = "http://your-server:8099"
 ```
 
-**2. Trust the plugin** (silences the security scan advisory)
+Omit this step if your server runs on `http://localhost:8099` (the default).
+
+**3. Trust the plugin** (silences the security scan advisory)
 
 ```bash
 openclaw config set plugins.allow '["agentmetrics"]'
 ```
 
-**3. Restart the gateway**
+**4. Restart the gateway**
 
 ```bash
 openclaw gateway restart
 ```
 
-**4. Verify**
+**5. Verify**
 
 ```bash
 openclaw plugins list
 # agentmetrics   loaded
 ```
 
+**6. Set your agent name** (recommended)
+
+In your agent's `openclaw.json`:
+
+```json
+{
+  "name": "my-agent"
+}
+```
+
+The `name` field becomes the agent ID in your dashboard. Give each agent a distinct name.
+
 ---
 
 ## What gets tracked
@@ -76,15 +99,15 @@ Every agent session reports automatically:
 
 ## Troubleshooting
 
-**"dangerous code patterns" warning on install**
-Safe to ignore. The plugin reads `AGENTMETRICS_API_KEY` and sends it as a Bearer token to the AgentMetrics API. Add `agentmetrics` to `plugins.allow` to suppress it permanently.
+**"dangerous code patterns" warning on install**  
+Safe to ignore. The plugin reads `AGENTMETRICS_BASE_URL` and makes network calls to the AgentMetrics API. Add `agentmetrics` to `plugins.allow` to suppress it permanently.
 
-**"manifest id does not match package name" warning**
+**"manifest id does not match package name" warning**  
 Not an error. The plugin's internal manifest id is `agentmetrics`; the npm package name is `agentmetrics-openclaw`. Use `agentmetrics` (not the npm name) in `plugins.allow`.
 
-**Runs not appearing in the dashboard**
-1. Check the key is set: `echo $AGENTMETRICS_API_KEY`
-2. Verify the plugin loads: `openclaw plugins list`
+**Runs not appearing in the dashboard**  
+1. Verify the plugin loads: `openclaw plugins list`
+2. Check the server is reachable: `openclaw agentmetrics test`
 3. Restart the gateway after any env var change
 4. Confirm your `openclaw.json` has a `name` field
 

package/index.ts

@@ -1,9 +1,99 @@
 import { randomUUID } from "crypto";
+import { appendFileSync, chmodSync, existsSync, mkdirSync, readFileSync, writeFileSync } from "fs";
+import { gzipSync } from "zlib";
+import { dirname, join } from "path";
+function _hashName(name: string): string {
+  let h = 0x811c9dc5;
+  for (let i = 0; i < name.length; i++) {
+    h = (Math.imul(h ^ name.charCodeAt(i), 0x01000193)) >>> 0;
+  }
+  return `t_${h.toString(16).padStart(8, "0")}`;
+}
+
+const _SECRET_PATTERNS: RegExp[] = [
+  /sk-[A-Za-z0-9\-_]{20,}/g,
+  /am_[A-Za-z0-9\-_]{16,}/g,
+  /\bey[A-Za-z0-9\-_]{20,}\.[A-Za-z0-9\-_]{20,}/g,
+  /(?:api[_\-]?key|apikey|api[_\-]?token|access[_\-]?token|secret|password|passwd|auth)[=:\s"']+([^\s"'&,\]}\n]{8,})/gi,
+];
+
+function _scrubSecrets(str: string): string {
+  let out = str;
+  for (const re of _SECRET_PATTERNS) {
+    out = out.replace(re, "[REDACTED]");
+  }
+  return out;
+}
+
+const _PRICING: Record<string, [number, number, number?, number?]> = {
+  "claude-opus-4-7":            [15.0,  75.0,  1.50, 18.75],
+  "claude-opus-4-5":            [15.0,  75.0,  1.50, 18.75],
+  "claude-opus-4":              [15.0,  75.0,  1.50, 18.75],
+  "claude-sonnet-4-6":          [ 3.0,  15.0,  0.30,  3.75],
+  "claude-sonnet-4-5":          [ 3.0,  15.0,  0.30,  3.75],
+  "claude-haiku-4-5":           [ 0.8,   4.0,  0.08,  1.00],
+  "claude-sonnet-3-7":          [ 3.0,  15.0,  0.30,  3.75],
+  "claude-3-5-sonnet-20241022": [ 3.0,  15.0,  0.30,  3.75],
+  "claude-3-5-sonnet-20240620": [ 3.0,  15.0,  0.30,  3.75],
+  "claude-3-5-haiku-20241022":  [ 0.8,   4.0,  0.08,  1.00],
+  "claude-3-opus-20240229":     [15.0,  75.0,  1.50, 18.75],
+  "claude-3-sonnet-20240229":   [ 3.0,  15.0],
+  "claude-3-haiku-20240307":    [ 0.25,  1.25, 0.03,  0.30],
+  "gpt-4o":                     [ 2.5,  10.0],
+  "gpt-4o-mini":                [ 0.15,  0.60],
+  "gpt-4-turbo":                [10.0,  30.0],
+  "gpt-4":                      [30.0,  60.0],
+  "gpt-3.5-turbo":              [ 0.50,  1.50],
+  "gemini-2.0-flash":           [ 0.075, 0.30],
+  "gemini-2.5-pro":             [ 1.25, 10.0],
+  "gemini-1.5-pro":             [ 1.25,  5.0],
+  "gemini-1.5-flash":           [ 0.075, 0.30],
+};
+
 
-// pluginConfig overrides env vars — set once in register()
 let API_KEY: string | undefined;
 let BASE_URL: string;
 
+let ENABLED              = true;
+let REDACTION_MODE: "strict" | "moderate" | "debug" = "strict";
+let EXPORTED_TOOL_NAMES: "allowlist" | "blocklist" | "hash" | "off" = "blocklist";
+let REDACT_TOOL_NAMES: string[] = [];
+let DEBUG_EXPIRES_AT: number | null = null;
+
+let FLUSH_INTERVAL_MS  = 10_000;
+let MAX_BATCH_SIZE      = 100;
+let MAX_QUEUE_SIZE      = 10_000;
+let RETRY_MAX_ATTEMPTS  = 5;
+let COMPRESS_PAYLOADS   = false;
+
+
+const _metrics = { sent: 0, failed: 0, dropped: 0 };
+
+
+interface QueuedEvent {
+  payload:    Record<string, unknown>;
+  attempt:    number;
+  enqueuedAt: number;
+}
+const _queue: QueuedEvent[] = [];
+const _dlq:   QueuedEvent[] = [];
+
+
+const CB_THRESHOLD = 10;
+const CB_PROBE_MS  = 5 * 60_000;
+type CbState = "closed" | "open" | "half-open";
+let _cbState: CbState = "closed";
+let _cbConsecFails    = 0;
+let _cbOpenAt: number | null = null;
+
+
+let WAL_PATH: string | null = null;
+let _flushTimer: ReturnType<typeof setInterval> | null = null;
+
+// Duplicate-registration guard
+let _registered = false;
+
+
 interface PluginApi {
   config:         Record<string, unknown>;
   pluginConfig?:  Record<string, unknown>;
@@ -11,12 +101,11 @@ interface PluginApi {
   registerCli?: (registrar: {
     name:        string;
     description: string;
-    commands:    Array<{ name: string; description: string; handler: () => void }>;
+    commands:    Array<{ name: string; description: string; handler: () => void | Promise<void> }>;
   }) => void;
   on: (hookName: string, handler: (...args: unknown[]) => void) => void;
 }
 
-// ─── Types (sourced from openclaw/src/plugins/hook-types.ts) ──────────────────
 
 type AgentContext = {
   runId?:           string;
@@ -151,7 +240,6 @@ type GatewayStartEvent = { port: number };
 type GatewayStopEvent  = { reason?: string };
 type GatewayContext    = { port?: number };
 
-// ─── Per-session and per-run state ────────────────────────────────────────────
 
 interface SessionMeta {
   traceId:     string;
@@ -159,6 +247,15 @@ interface SessionMeta {
   startedAt:   number;
   compactions: number;
   resets:      number;
+  // Session-level aggregates accumulated across all runs in this session
+  runCount:              number;
+  totalInputTokens:      number;
+  totalOutputTokens:     number;
+  totalCacheReadTokens:  number;
+  totalCacheWriteTokens: number;
+  totalToolCalls:        number;
+  totalEstimatedCostUsd: number;
+  totalDurationMs:       number;
 }
 
 interface RunMeta {
@@ -176,43 +273,201 @@ interface RunMeta {
   model?:           string;
   provider?:        string;
   sessionKey?:      string;
-  startedAt:        number; // run-level start time for accurate duration
+  startedAt:        number;
 }
 
 const sessions = new Map<string, SessionMeta>();
 const runs     = new Map<string, RunMeta>();
 
-// ─── HTTP helpers ─────────────────────────────────────────────────────────────
 
-/** Send a completed run summary to /v1/events (persisted to DB). */
-async function send(payload: Record<string, unknown>): Promise<void> {
-  if (!API_KEY) return;
+function _walAppend(payload: Record<string, unknown>): void {
+  if (!WAL_PATH) return;
   try {
-    await fetch(`${BASE_URL}/v1/events`, {
-      method: "POST",
-      headers: {
-        "Content-Type":  "application/json",
-        "Authorization": `Bearer ${API_KEY}`,
-      },
-      body: JSON.stringify(payload),
+    appendFileSync(WAL_PATH, JSON.stringify(payload) + "\n", "utf8");
+  } catch { /* non-fatal */ }
+}
+
+function _walCompact(sentIds: Set<string>): void {
+  if (!WAL_PATH || !existsSync(WAL_PATH)) return;
+  try {
+    const lines = readFileSync(WAL_PATH, "utf8").split("\n").filter(Boolean);
+    const kept  = lines.filter((line) => {
+      try {
+        const ev = JSON.parse(line) as Record<string, unknown>;
+        return !sentIds.has(String(ev.event_id ?? ""));
+      } catch { return false; }
     });
+    writeFileSync(WAL_PATH, kept.length ? kept.join("\n") + "\n" : "", "utf8");
+  } catch { /* non-fatal */ }
+}
+
+function _walRecover(): void {
+  if (!WAL_PATH || !existsSync(WAL_PATH)) return;
+  try {
+    const lines = readFileSync(WAL_PATH, "utf8").split("\n").filter(Boolean);
+    for (const line of lines) {
+      try {
+        const payload = JSON.parse(line) as Record<string, unknown>;
+        _enqueue(payload, true);
+      } catch { /* skip corrupt lines */ }
+    }
+    if (lines.length > 0) {
+      console.log(`AgentMetrics: recovered ${lines.length} event(s) from WAL`);
+    }
+  } catch { /* non-fatal */ }
+}
+
+
+function _enqueue(payload: Record<string, unknown>, fromWal = false): void {
+  if (_queue.length >= MAX_QUEUE_SIZE) {
+    _queue.shift(); // FIFO: drop oldest on overflow
+    _metrics.dropped += 1;
+  }
+  _queue.push({ payload, attempt: 0, enqueuedAt: Date.now() });
+  if (!fromWal) _walAppend(payload);
+}
+
+
+function _cbIsOpen(): boolean {
+  if (_cbState === "closed") return false;
+  if (_cbState === "open") {
+    if (_cbOpenAt !== null && Date.now() - _cbOpenAt >= CB_PROBE_MS) {
+      _cbState = "half-open";
+      return false; // let one probe through
+    }
+    return true;
+  }
+  return false; // half-open: probe is allowed
+}
+
+function _cbOnSuccess(): void {
+  if (_cbState !== "closed") {
+    console.log("AgentMetrics: circuit breaker closed - delivery resumed");
+  }
+  _cbState       = "closed";
+  _cbConsecFails = 0;
+  _cbOpenAt      = null;
+}
+
+function _cbOnFailure(): void {
+  _cbConsecFails += 1;
+  if (_cbState === "half-open" || _cbConsecFails >= CB_THRESHOLD) {
+    _cbState  = "open";
+    _cbOpenAt = Date.now();
+    console.log(
+      `AgentMetrics: circuit breaker opened after ${_cbConsecFails} consecutive failures - ` +
+      `probing again in ${CB_PROBE_MS / 60_000}min`,
+    );
+  }
+}
+
+
+function _buildHeaders(): Record<string, string> {
+  return {
+    "Content-Type":  "application/json",
+    "Authorization": `Bearer ${API_KEY}`,
+  };
+}
+
+function _maybeGzip(body: string): { body: string | Uint8Array; extra: Record<string, string> } {
+  if (COMPRESS_PAYLOADS && body.length > 1024) {
+    try {
+      return {
+        body:  gzipSync(Buffer.from(body, "utf8")),
+        extra: { "Content-Encoding": "gzip" },
+      };
+    } catch { /* fall through */ }
+  }
+  return { body, extra: {} };
+}
+
+function _requeue(batch: QueuedEvent[]): void {
+  for (const item of batch) {
+    item.attempt  += 1;
+    _metrics.failed += 1;
+    if (item.attempt >= RETRY_MAX_ATTEMPTS) {
+      _dlq.push(item);
+    } else {
+      _queue.push(item); // back of queue, not front - prevent starvation
+    }
+  }
+}
+
+async function _flushBatch(batch: QueuedEvent[]): Promise<void> {
+  const rawBody     = JSON.stringify({ events: batch.map((e) => e.payload) });
+  const { body, extra } = _maybeGzip(rawBody);
+  try {
+    const resp = await fetch(`${BASE_URL}/v1/events/batch`, {
+      method:  "POST",
+      headers: { ..._buildHeaders(), ...extra },
+      body,
+    });
+    if (resp.ok) {
+      _cbOnSuccess();
+      _metrics.sent += batch.length;
+      _walCompact(new Set(batch.map((e) => String(e.payload.event_id ?? ""))));
+    } else if (resp.status === 404) {
+      await _flushIndividual(batch); // batch endpoint not yet available
+    } else {
+      _cbOnFailure();
+      _requeue(batch);
+    }
   } catch {
-    // Never crash the agent on observability failure
+    _cbOnFailure();
+    _requeue(batch);
+  }
+}
+
+async function _flushIndividual(batch: QueuedEvent[]): Promise<void> {
+  for (const item of batch) {
+    const rawBody         = JSON.stringify(item.payload);
+    const { body, extra } = _maybeGzip(rawBody);
+    try {
+      const resp = await fetch(`${BASE_URL}/v1/events`, {
+        method:  "POST",
+        headers: { ..._buildHeaders(), ...extra },
+        body,
+      });
+      if (resp.ok) {
+        _cbOnSuccess();
+        _metrics.sent += 1;
+        _walCompact(new Set([String(item.payload.event_id ?? "")]));
+      } else {
+        _cbOnFailure();
+        _requeue([item]);
+      }
+    } catch {
+      _cbOnFailure();
+      _requeue([item]);
+    }
   }
 }
 
+async function _flush(): Promise<void> {
+  if (!API_KEY || _queue.length === 0 || _cbIsOpen()) return;
+  const batch = _queue.splice(0, MAX_BATCH_SIZE);
+  await _flushBatch(batch);
+}
+
+
+/** Enqueue a completed run summary; delivered by the flush loop. */
+function send(payload: Record<string, unknown>): void {
+  if (!API_KEY || !ENABLED) return;
+  _enqueue(payload);
+}
+
 /**
  * Send a real-time intermediate event to /v1/activity (in-memory, streamed
- * to dashboard via SSE). Fire-and-forget — no retry, no crash on failure.
+ * to dashboard via SSE). Fire-and-forget - no retry, no crash on failure.
  */
 function sendActivity(
-  type:        string,
-  agentId:     string,
-  sessionKey:  string | undefined,
-  runId:       string | undefined,
-  data?:       Record<string, unknown>,
+  type:       string,
+  agentId:    string,
+  sessionKey: string | undefined,
+  runId:      string | undefined,
+  data?:      Record<string, unknown>,
 ): void {
-  if (!API_KEY) return;
+  if (!API_KEY || !ENABLED) return;
   const payload = JSON.stringify({
     type,
     agent_id:    agentId,
@@ -243,12 +498,73 @@ function emptyRun(sessionKey?: string): RunMeta {
   };
 }
 
-// ─── Plugin ───────────────────────────────────────────────────────────────────
+
+
+function _activeMode(): "strict" | "moderate" | "debug" {
+  if (DEBUG_EXPIRES_AT !== null) {
+    if (Date.now() < DEBUG_EXPIRES_AT) return "debug";
+    DEBUG_EXPIRES_AT = null;
+    console.log("AgentMetrics: debug mode expired - reverting to strict redaction");
+  }
+  return REDACTION_MODE;
+}
+
+function _redactError(err: string | undefined, activityStream = false): string | undefined {
+  if (!err) return err;
+  const mode = _activeMode();
+  if (mode === "debug") return err;
+  const maxLen = (mode === "strict" || activityStream) ? 200 : 500;
+  return _scrubSecrets(err).slice(0, maxLen);
+}
+
+function _redactToolName(name: string): string {
+  const mode = _activeMode();
+  if (mode === "debug") return name;
+  // allowlist mode: redactToolNames is the permitted list; blocklist mode: it's the deny list
+  switch (EXPORTED_TOOL_NAMES) {
+    case "off":
+      return "[REDACTED]";
+    case "hash":
+      return _hashName(name);
+    case "allowlist":
+      return REDACT_TOOL_NAMES.includes(name) ? name : `[REDACTED:${_hashName(name)}]`;
+    case "blocklist":
+    default:
+      return REDACT_TOOL_NAMES.includes(name) ? `[REDACTED:${_hashName(name)}]` : name;
+  }
+}
+
+function _redactToolNames(names: string[]): string[] {
+  return names.map(_redactToolName);
+}
+
+
+
+function _estimateCost(
+  model:           string | undefined,
+  input:           number,
+  output:          number,
+  cacheRead:       number,
+  cacheWrite:      number,
+): number | undefined {
+  if (!model) return undefined;
+  // Try exact match first, then strip date suffix (e.g. -20241022)
+  const rates = _PRICING[model.toLowerCase()] ?? _PRICING[model.toLowerCase().replace(/-\d{8}$/, "")];
+  if (!rates) return undefined;
+  const M = 1_000_000;
+  return (
+    input      * rates[0]          / M +
+    output     * rates[1]          / M +
+    cacheRead  * (rates[2] ?? 0)   / M +
+    cacheWrite * (rates[3] ?? 0)   / M
+  );
+}
+
 
 const plugin = {
   id:          "agentmetrics",
   name:        "AgentMetrics",
-  description: "360-degree observability for every OpenClaw agent — real-time streaming, tokens, tools, latency, cost, subagents, and reliability.",
+  description: "360-degree observability for every OpenClaw agent - real-time streaming, tokens, tools, latency, cost, subagents, and reliability.",
   configSchema: {
     type: "object",
     properties: {
@@ -258,41 +574,125 @@ const plugin = {
       },
       endpoint: {
         type:        "string",
-        description: "Custom API endpoint (default: https://api.agentmetrics.dev)",
+        description: "Custom API endpoint (default: http://localhost:8099)",
+      },
+      enabled: {
+        type:        "boolean",
+        description: "Disable the plugin without removing it (default: true)",
+      },
+      flushIntervalSeconds: {
+        type:        "number",
+        description: "How often to flush the event queue to the API (default: 10)",
+      },
+      maxBatchSize: {
+        type:        "number",
+        description: "Maximum events per batch request (default: 100)",
+      },
+      maxQueueSize: {
+        type:        "number",
+        description: "Maximum in-memory queue depth before FIFO drop (default: 10000)",
+      },
+      retryMaxAttempts: {
+        type:        "number",
+        description: "Max retry attempts before moving event to DLQ (default: 5)",
+      },
+      redactionMode: {
+        type:        "string",
+        enum:        ["strict", "moderate", "debug"],
+        description: "PII redaction level applied to prompts/completions (default: strict)",
+      },
+      exportedToolNames: {
+        type:        "string",
+        enum:        ["allowlist", "blocklist", "hash", "off"],
+        description: "Which tool names to include in exports (default: blocklist)",
+      },
+      redactToolNames: {
+        type:        "array",
+        items:       { type: "string" },
+        description: "Tool names to redact when exportedToolNames is 'blocklist'",
+      },
+      compressPayloads: {
+        type:        "boolean",
+        description: "Gzip-compress batch payloads larger than 1 KB (default: false)",
       },
     },
     additionalProperties: false,
   } as const,
 
   register(api: PluginApi) {
-    // Config: pluginConfig > env vars
+    if (_registered) {
+      console.warn(
+        "\n  AgentMetrics: ⚠ register() called twice - possible duplicate instrumentation.\n" +
+        "  If you have both the plugin and an SDK hook active, remove one to avoid\n" +
+        "  double-counting runs and inflated token/cost totals.\n",
+      );
+    }
+    _registered = true;
+
     API_KEY  = (api.pluginConfig?.apiKey  as string | undefined) ?? process.env.AGENTMETRICS_API_KEY;
     BASE_URL = (
       (api.pluginConfig?.endpoint as string | undefined) ??
       process.env.AGENTMETRICS_URL ??
-      "https://api.agentmetrics.dev"
+      "http://localhost:8099"
     ).replace(/\/$/, "");
 
-    // Auto-enable when a key is available (env var or plugin config)
+    ENABLED            = (api.pluginConfig?.enabled           as boolean       | undefined) ?? true;
+    REDACTION_MODE     = (api.pluginConfig?.redactionMode      as typeof REDACTION_MODE    | undefined) ?? "strict";
+    EXPORTED_TOOL_NAMES   = (api.pluginConfig?.exportedToolNames   as typeof EXPORTED_TOOL_NAMES  | undefined) ?? "blocklist";
+    REDACT_TOOL_NAMES  = (api.pluginConfig?.redactToolNames    as string[]      | undefined) ?? [];
+    FLUSH_INTERVAL_MS  = ((api.pluginConfig?.flushIntervalSeconds as number | undefined) ?? 10) * 1000;
+    MAX_BATCH_SIZE     = (api.pluginConfig?.maxBatchSize       as number        | undefined) ?? 100;
+    MAX_QUEUE_SIZE     = (api.pluginConfig?.maxQueueSize       as number        | undefined) ?? 10_000;
+    RETRY_MAX_ATTEMPTS = (api.pluginConfig?.retryMaxAttempts   as number        | undefined) ?? 5;
+    COMPRESS_PAYLOADS  = (api.pluginConfig?.compressPayloads   as boolean       | undefined) ?? false;
+
+    if (REDACTION_MODE === "debug") {
+      DEBUG_EXPIRES_AT = Date.now() + 60 * 60 * 1000;
+      console.log("AgentMetrics: ⚠ debug redaction mode active - expires in 1 hour");
+    }
+
     if (typeof api.registerAutoEnableProbe === "function") {
-      api.registerAutoEnableProbe(() => !!API_KEY);
+      api.registerAutoEnableProbe(() => !!API_KEY && ENABLED);
+    }
+
+    if (!ENABLED) {
+      console.log("\n  AgentMetrics: disabled via config (metrics.enabled: false)\n");
+      return;
     }
 
     if (!API_KEY) {
       console.log(
         "\n  AgentMetrics: no API key found.\n" +
         "  Your agent runs are not being tracked.\n" +
-        "  Get an API key at https://agentmetrics.dev and set AGENTMETRICS_API_KEY.\n"
+        "  Start AgentMetrics (see README) and set AGENTMETRICS_API_KEY.\n" +
+        "  AGENTMETRICS_URL defaults to http://localhost:8099.\n",
       );
       return;
     }
 
+    try {
+      const home = process.env.HOME ?? process.env.USERPROFILE ?? process.cwd();
+      WAL_PATH   = join(home, ".config", "openclaw", "agentmetrics-wal.jsonl");
+      mkdirSync(dirname(WAL_PATH), { recursive: true });
+      try { chmodSync(dirname(WAL_PATH), 0o700); } catch {}
+      _walRecover();
+    } catch {
+      WAL_PATH = null; // WAL unavailable - queue still works in-memory
+    }
+
+    if (_flushTimer) clearInterval(_flushTimer);
+    _flushTimer = setInterval(() => { _flush().catch(() => {}); }, FLUSH_INTERVAL_MS);
+    // Allow process to exit even with an active timer
+    if (typeof (_flushTimer as unknown as { unref?: () => void }).unref === "function") {
+      (_flushTimer as unknown as { unref: () => void }).unref();
+    }
+
     console.log(
-      `\n  AgentMetrics active — sending data to ${BASE_URL}\n` +
-      `  View your dashboard → https://agentmetrics.dev\n`
+      `\n  AgentMetrics active - sending data to ${BASE_URL}\n` +
+      `  Queue: max ${MAX_QUEUE_SIZE} events, batch ${MAX_BATCH_SIZE}, flush every ${FLUSH_INTERVAL_MS / 1000}s\n` +
+      `  View your dashboard → http://localhost:3099\n`,
     );
 
-    // CLI: openclaw agentmetrics status
     if (typeof api.registerCli === "function") {
       api.registerCli({
         name:        "agentmetrics",
@@ -300,39 +700,182 @@ const plugin = {
         commands: [
           {
             name:        "status",
-            description: "Show current AgentMetrics plugin status",
+            description: "Show current plugin status, config, delivery counters, and circuit breaker state",
             handler() {
               const keyPreview = API_KEY
                 ? `${API_KEY.slice(0, 8)}...${API_KEY.slice(-4)}`
                 : "(not set)";
-              console.log("AgentMetrics — active");
-              console.log(`  API key  : ${keyPreview}`);
-              console.log(`  Endpoint : ${BASE_URL}`);
-              console.log(`  Sessions : ${sessions.size} tracked`);
-              console.log(`  Runs     : ${runs.size} in flight`);
+              const mode = _activeMode();
+              const cbInfo = _cbState === "open" && _cbOpenAt
+                ? ` (opens probe at ${new Date(_cbOpenAt + CB_PROBE_MS).toLocaleTimeString()})`
+                : "";
+              console.log("AgentMetrics - status");
+              console.log(`  API key          : ${keyPreview}`);
+              console.log(`  Endpoint         : ${BASE_URL}`);
+              console.log(`  Redaction        : ${mode}${mode === "debug" && DEBUG_EXPIRES_AT ? ` (expires ${new Date(DEBUG_EXPIRES_AT).toLocaleTimeString()})` : ""}`);
+              console.log(`  Tool names       : ${EXPORTED_TOOL_NAMES}`);
+              console.log(`  Compress payloads: ${COMPRESS_PAYLOADS}`);
+              console.log(`  Flush interval   : ${FLUSH_INTERVAL_MS / 1000}s`);
+              console.log(`  WAL path         : ${WAL_PATH ?? "(unavailable)"}`);
+              console.log("");
+              console.log(`  Circuit breaker  : ${_cbState}${cbInfo}`);
+              console.log(`  Queue depth      : ${_queue.length} / ${MAX_QUEUE_SIZE}`);
+              console.log(`  DLQ depth        : ${_dlq.length}`);
+              console.log(`  Sessions tracked : ${sessions.size}`);
+              console.log(`  Runs in flight   : ${runs.size}`);
+              console.log("");
+              console.log(`  Sent             : ${_metrics.sent}`);
+              console.log(`  Failed           : ${_metrics.failed}`);
+              console.log(`  Dropped (overflow): ${_metrics.dropped}`);
+            },
+          },
+          {
+            name:        "flush",
+            description: "Force-flush all queued events immediately",
+            async handler() {
+              const before = _queue.length;
+              if (before === 0) {
+                console.log("AgentMetrics flush - queue empty, nothing to flush");
+                return;
+              }
+              if (_cbIsOpen()) {
+                console.log(`AgentMetrics flush - circuit breaker is ${_cbState}, skipping`);
+                return;
+              }
+              console.log(`AgentMetrics flush - flushing ${before} event(s)…`);
+              // Drain the entire queue in batches
+              while (_queue.length > 0 && !_cbIsOpen()) {
+                await _flush();
+              }
+              console.log(`  Done - sent: ${_metrics.sent}, failed: ${_metrics.failed}, queued: ${_queue.length}`);
+            },
+          },
+          {
+            name:        "tail",
+            description: "Show recent in-flight run state",
+            handler() {
+              if (runs.size === 0 && sessions.size === 0) {
+                console.log("AgentMetrics tail - no active sessions or runs");
+                return;
+              }
+              console.log("AgentMetrics tail - active state");
+              if (sessions.size > 0) {
+                console.log(`  Sessions (${sessions.size}):`);
+                for (const [key, s] of sessions) {
+                  console.log(`    ${key.slice(0, 12)}… agent=${s.agentId} compactions=${s.compactions} resets=${s.resets}`);
+                }
+              }
+              if (runs.size > 0) {
+                console.log(`  Runs in flight (${runs.size}):`);
+                for (const [id, r] of runs) {
+                  const age = Math.round((Date.now() - r.startedAt) / 1000);
+                  console.log(`    ${id.slice(0, 12)}… llm=${r.llmCalls} tools=${r.toolCalls} ${age}s elapsed`);
+                }
+              }
+            },
+          },
+          {
+            name:        "test",
+            description: "Send a test event and verify end-to-end delivery",
+            async handler() {
+              if (!API_KEY) {
+                console.log("AgentMetrics test - no API key set, cannot send");
+                return;
+              }
+              console.log(`AgentMetrics test - sending to ${BASE_URL}…`);
+              try {
+                const resp = await fetch(`${BASE_URL}/v1/events`, {
+                  method: "POST",
+                  headers: {
+                    "Content-Type":  "application/json",
+                    "Authorization": `Bearer ${API_KEY}`,
+                  },
+                  body: JSON.stringify({
+                    event_id:                 randomUUID(),
+                    trace_id:                 randomUUID(),
+                    agent_id:                 "agentmetrics-test",
+                    platform:                 "openclaw",
+                    event_name:               "agent_end",
+                    ts:                       Date.now(),
+                    status:                   "success",
+                    duration_ms:              1,
+                    redaction_policy_version: `v1-${_activeMode()}`,
+                  }),
+                });
+                if (resp.ok) {
+                  console.log(`  ✓ Delivered - HTTP ${resp.status}`);
+                } else {
+                  const body = await resp.text().catch(() => "");
+                  console.log(`  ✗ Failed - HTTP ${resp.status} ${body.slice(0, 200)}`);
+                }
+              } catch (err) {
+                console.log(`  ✗ Failed - ${err}`);
+              }
+            },
+          },
+          {
+            name:        "redaction-check",
+            description: "Show what the current redaction policy does to a sample payload",
+            handler() {
+              const mode = _activeMode();
+              const sampleError   = "Connection failed: Bearer sk-ant-abc123exampletoken and api_key=supersecret";
+              const sampleTools   = ["bash", "read_file", "write_file", "send_email"];
+              const redactedError = _redactError(sampleError);
+              const redactedTools = _redactToolNames(sampleTools);
+              console.log("AgentMetrics redaction-check");
+              console.log(`  Mode          : ${mode}`);
+              console.log(`  Tool export   : ${EXPORTED_TOOL_NAMES}`);
+              console.log(`  Blocked names : ${REDACT_TOOL_NAMES.length ? REDACT_TOOL_NAMES.join(", ") : "(none)"}`);
+              console.log("");
+              console.log("  Error sample:");
+              console.log(`    Input  : ${sampleError}`);
+              console.log(`    Output : ${redactedError}`);
+              console.log("");
+              console.log("  Tool name sample:");
+              sampleTools.forEach((t, i) =>
+                console.log(`    ${t.padEnd(16)} → ${redactedTools[i]}`),
+              );
+            },
+          },
+          {
+            name:        "drain",
+            description: "Retry all events in the dead-letter queue",
+            async handler() {
+              if (_dlq.length === 0) {
+                console.log("AgentMetrics drain - DLQ is empty");
+                return;
+              }
+              const count = _dlq.length;
+              console.log(`AgentMetrics drain - retrying ${count} DLQ event(s)…`);
+              // Reset attempts and move DLQ back to main queue
+              const items = _dlq.splice(0, _dlq.length);
+              for (const item of items) {
+                item.attempt = 0;
+                _queue.push(item);
+              }
+              // Flush immediately
+              while (_queue.length > 0 && !_cbIsOpen()) {
+                await _flush();
+              }
+              console.log(`  Done - sent: ${_metrics.sent}, failed: ${_metrics.failed}, remaining DLQ: ${_dlq.length}`);
             },
           },
         ],
       });
     }
 
-    // ── gateway_start ─────────────────────────────────────────────────────
-    // Activity only — gateway events are infrastructure, not agent runs.
-    // We do NOT call send() here to avoid polluting the agents list.
     api.on("gateway_start", (event: GatewayStartEvent, _ctx: GatewayContext) => {
       sendActivity("gateway_start", "openclaw-gateway", undefined, undefined, {
         port: event.port,
       });
     });
 
-    // ── gateway_stop ──────────────────────────────────────────────────────
     api.on("gateway_stop", (event: GatewayStopEvent, _ctx: GatewayContext) => {
       sendActivity("gateway_stop", "openclaw-gateway", undefined, undefined, {
         reason: event.reason,
       });
     });
 
-    // ── session_start ─────────────────────────────────────────────────────
     api.on("session_start", (event: SessionStartEvent, ctx: SessionContext) => {
       const key = event.sessionKey ?? event.sessionId;
       if (!key || sessions.has(key)) return;
@@ -343,11 +886,17 @@ const plugin = {
         startedAt:   Date.now(),
         compactions: 0,
         resets:      0,
+        runCount:              0,
+        totalInputTokens:      0,
+        totalOutputTokens:     0,
+        totalCacheReadTokens:  0,
+        totalCacheWriteTokens: 0,
+        totalToolCalls:        0,
+        totalEstimatedCostUsd: 0,
+        totalDurationMs:       0,
       });
     });
 
-    // ── before_agent_start ────────────────────────────────────────────────
-    // Fires at the start of each run. Sends run_start for real-time UI.
     api.on("before_agent_start", (event: BeforeAgentStartEvent, ctx: AgentContext) => {
       const sessionKey = ctx.sessionKey ?? ctx.sessionId;
       const agentId    = ctx.agentId ?? "openclaw-agent";
@@ -361,9 +910,6 @@ const plugin = {
       });
     });
 
-    // ── llm_input ─────────────────────────────────────────────────────────
-    // Fires before each LLM call. Sends llm_start for real-time "thinking"
-    // indicator. Also accumulates LLM call count and image count.
     api.on("llm_input", (event: LlmInputEvent, ctx: AgentContext) => {
       const { runId } = event;
       if (!runId) return;
@@ -387,9 +933,6 @@ const plugin = {
       });
     });
 
-    // ── llm_output ────────────────────────────────────────────────────────
-    // Fires after each LLM response. Accumulates tokens and sends llm_end
-    // so the dashboard can show token counts updating in real time.
     api.on("llm_output", (event: LlmOutputEvent, ctx: AgentContext) => {
       const { runId } = event;
       if (!runId) return;
@@ -421,9 +964,6 @@ const plugin = {
       });
     });
 
-    // ── before_tool_call ──────────────────────────────────────────────────
-    // Fires before each tool execution. Sends tool_start immediately so
-    // the live visualizer shows tool activity in real time.
     api.on("before_tool_call", (event: BeforeToolCallEvent, ctx: ToolContext) => {
       const runId      = event.runId ?? ctx.runId;
       const sessionKey = ctx.sessionKey ?? ctx.sessionId;
@@ -434,9 +974,6 @@ const plugin = {
       });
     });
 
-    // ── after_tool_call ───────────────────────────────────────────────────
-    // Fires after every tool execution. Counts calls/errors and sends
-    // tool_end with outcome and duration.
     api.on("after_tool_call", (event: AfterToolCallEvent, ctx: ToolContext) => {
       const runId      = event.runId ?? ctx.runId;
       const sessionKey = ctx.sessionKey ?? ctx.sessionId;
@@ -452,13 +989,12 @@ const plugin = {
       }
 
       sendActivity("tool_end", agentId, sessionKey, runId, {
-        tool_name:   event.toolName,
+        tool_name:   _redactToolName(event.toolName),
         duration_ms: event.durationMs,
-        error:       event.error?.slice(0, 200),
+        error:       _redactError(event.error, true),
       });
     });
 
-    // ── subagent_spawning ─────────────────────────────────────────────────
     api.on("subagent_spawning", (event: SubagentSpawningEvent, ctx: SubagentContext) => {
       const runId      = ctx.runId;
       const sessionKey = ctx.requesterSessionKey;
@@ -480,7 +1016,6 @@ const plugin = {
       });
     });
 
-    // ── subagent_ended ────────────────────────────────────────────────────
     api.on("subagent_ended", (event: SubagentEndedEvent, ctx: SubagentContext) => {
       const runId      = event.runId ?? ctx.runId;
       const sessionKey = ctx.requesterSessionKey;
@@ -497,11 +1032,10 @@ const plugin = {
       sendActivity("subagent_end", agentId, sessionKey, runId, {
         child_session_key: event.targetSessionKey,
         outcome:           event.outcome,
-        error:             event.error?.slice(0, 200),
+        error:             _redactError(event.error, true),
       });
     });
 
-    // ── before_compaction ─────────────────────────────────────────────────
     api.on("before_compaction", (_event: CompactionEvent, ctx: AgentContext) => {
       const key     = ctx.sessionKey ?? ctx.sessionId;
       const agentId = ctx.agentId ?? sessions.get(key ?? "")?.agentId ?? "openclaw-agent";
@@ -516,7 +1050,6 @@ const plugin = {
       sendActivity("compaction", agentId, key, ctx.runId);
     });
 
-    // ── before_reset ──────────────────────────────────────────────────────
     api.on("before_reset", (event: ResetEvent, ctx: AgentContext) => {
       const key     = ctx.sessionKey ?? ctx.sessionId;
       const agentId = ctx.agentId ?? sessions.get(key ?? "")?.agentId ?? "openclaw-agent";
@@ -531,9 +1064,6 @@ const plugin = {
       sendActivity("reset", agentId, key, ctx.runId, { reason: event.reason });
     });
 
-    // ── agent_end ─────────────────────────────────────────────────────────
-    // Fires when the agent finishes a run. Sends run_end for real-time UI,
-    // then the full persisted event to /v1/events with all accumulated data.
     api.on("agent_end", (event: AgentEndEvent, ctx: AgentContext) => {
       const sessionKey = ctx.sessionKey ?? ctx.sessionId;
       if (!sessionKey) return;
@@ -550,22 +1080,47 @@ const plugin = {
         (run?.cacheReadTokens  ?? 0) +
         (run?.cacheWriteTokens ?? 0);
 
-      // Use run-level duration (accurate) — fall back to session duration only if unavailable
-      const durationMs = event.durationMs ?? (run ? Date.now() - run.startedAt : undefined);
+      const durationMs    = event.durationMs ?? (run ? Date.now() - run.startedAt : undefined);
+      const redactedError = _redactError(event.error);
+
+      const estimatedCostUsd = _estimateCost(
+        run?.model,
+        run?.inputTokens      ?? 0,
+        run?.outputTokens     ?? 0,
+        run?.cacheReadTokens  ?? 0,
+        run?.cacheWriteTokens ?? 0,
+      );
+
+      // Accumulate into session-level totals
+      if (session) {
+        session.runCount              += 1;
+        session.totalInputTokens      += run?.inputTokens      ?? 0;
+        session.totalOutputTokens     += run?.outputTokens     ?? 0;
+        session.totalCacheReadTokens  += run?.cacheReadTokens  ?? 0;
+        session.totalCacheWriteTokens += run?.cacheWriteTokens ?? 0;
+        session.totalToolCalls        += run?.toolCalls        ?? 0;
+        session.totalEstimatedCostUsd += estimatedCostUsd      ?? 0;
+        session.totalDurationMs       += durationMs            ?? 0;
+      }
 
-      // Real-time: notify dashboard the run finished
       sendActivity("run_end", agentId, sessionKey, ctx.runId, {
         status:       event.success ? "success" : "failed",
         duration_ms:  durationMs,
         total_tokens: totalTokens || undefined,
         tool_calls:   run?.toolCalls,
-        error:        event.error?.slice(0, 200),
+        error:        _redactError(event.error, true),
       });
 
-      // Persistent: full summary stored in DB
       send({
-        trace_id:            session?.traceId ?? randomUUID(),
-        agent_id:            agentId,
+        event_id:                 randomUUID(),
+        trace_id:                 session?.traceId ?? randomUUID(),
+        session_id:               sessionKey,
+        run_id:                   ctx.runId,
+        agent_id:                 agentId,
+        platform:                 "openclaw",
+        event_name:               "agent_end",
+        ts:                       Date.now(),
+        redaction_policy_version: `v1-${_activeMode()}`,
         status:              event.success ? "success" : "failed",
         duration_ms:         durationMs,
         model:               run?.model,
@@ -577,8 +1132,10 @@ const plugin = {
         total_tokens:        totalTokens || undefined,
         tool_calls:          run?.toolCalls  ?? 0,
         tool_errors:         run?.toolErrors ?? 0,
-        step_count:          event.messages?.length,
-        ...(event.error ? { error: event.error.slice(0, 500) } : {}),
+        tool_names:           run ? _redactToolNames([...run.toolNames]) : [],
+        step_count:           event.messages?.length,
+        ...(estimatedCostUsd != null ? { estimated_cost_usd: estimatedCostUsd } : {}),
+        ...(redactedError ? { error: redactedError } : {}),
         metadata: {
           llm_calls:         run?.llmCalls         ?? 0,
           images_count:      run?.imagesCount       ?? 0,
@@ -586,16 +1143,52 @@ const plugin = {
           subagent_errors:   run?.subagentErrors    ?? 0,
           compactions:       session?.compactions   ?? 0,
           resets:            session?.resets        ?? 0,
-          tool_names:        run ? [...run.toolNames] : [],
         },
       });
 
       if (ctx.runId) runs.delete(ctx.runId);
     });
 
-    // ── session_end ───────────────────────────────────────────────────────
     api.on("session_end", (event: SessionEndEvent, _ctx: SessionContext) => {
-      const key = event.sessionKey ?? event.sessionId;
+      const key     = event.sessionKey ?? event.sessionId;
+      const session = sessions.get(key);
+
+      if (session && session.runCount > 0) {
+        const sessionDurationMs = event.durationMs ?? (Date.now() - session.startedAt);
+        const totalTokens       =
+          session.totalInputTokens + session.totalOutputTokens +
+          session.totalCacheReadTokens + session.totalCacheWriteTokens;
+
+        send({
+          event_id:                 randomUUID(),
+          trace_id:                 session.traceId,
+          session_id:               key,
+          agent_id:                 session.agentId,
+          platform:                 "openclaw",
+          event_name:               "session_metrics",
+          ts:                       Date.now(),
+          redaction_policy_version: `v1-${_activeMode()}`,
+          status:                   "success",
+          duration_ms:              sessionDurationMs,
+          input_tokens:             session.totalInputTokens,
+          output_tokens:            session.totalOutputTokens,
+          cache_read_tokens:        session.totalCacheReadTokens,
+          cache_write_tokens:       session.totalCacheWriteTokens,
+          total_tokens:             totalTokens || undefined,
+          tool_calls:               session.totalToolCalls,
+          ...(session.totalEstimatedCostUsd > 0
+            ? { estimated_cost_usd: session.totalEstimatedCostUsd }
+            : {}),
+          metadata: {
+            run_count:     session.runCount,
+            compactions:   session.compactions,
+            resets:        session.resets,
+            message_count: event.messageCount,
+            reason:        event.reason,
+          },
+        });
+      }
+
       sessions.delete(key);
     });
   },

package/openclaw.plugin.json

@@ -1,7 +1,7 @@
 {
   "id": "agentmetrics",
   "name": "AgentMetrics",
-  "description": "360-degree observability for every OpenClaw agent — tokens, tools, latency, cost, subagents, context health, and reliability.",
+  "description": "360-degree observability for every OpenClaw agent - tokens, tools, latency, cost, subagents, context health, and reliability.",
   "configSchema": {
     "type": "object",
     "properties": {
@@ -11,7 +11,55 @@
       },
       "endpoint": {
         "type": "string",
-        "description": "Custom API endpoint (default: https://api.agentmetrics.dev)"
+        "description": "Custom API endpoint (default: http://localhost:8099)"
+      },
+      "enabled": {
+        "type": "boolean",
+        "description": "Master toggle - set to false to disable all telemetry without uninstalling (default: true)",
+        "default": true
+      },
+      "flushIntervalSeconds": {
+        "type": "integer",
+        "description": "How often to flush the event batch to the API in seconds (default: 10)",
+        "default": 10
+      },
+      "maxBatchSize": {
+        "type": "integer",
+        "description": "Maximum number of events per batch request (default: 100)",
+        "default": 100
+      },
+      "maxQueueSize": {
+        "type": "integer",
+        "description": "Maximum number of events to hold in the local queue before dropping (FIFO, default: 10000)",
+        "default": 10000
+      },
+      "retryMaxAttempts": {
+        "type": "integer",
+        "description": "Maximum retry attempts per event batch before moving to dead-letter queue (default: 5)",
+        "default": 5
+      },
+      "redactionMode": {
+        "type": "string",
+        "enum": ["strict", "moderate", "debug"],
+        "description": "Redaction policy: strict = all sensitive fields scrubbed (default); moderate = errors and secrets only; debug = opt-in full detail, auto-expires after 1h",
+        "default": "strict"
+      },
+      "toolNameExport": {
+        "type": "string",
+        "enum": ["allowlist", "blocklist", "hash", "off"],
+        "description": "How tool names are exported: blocklist = export all except redactToolNames (default); allowlist = export only listed; hash = pseudonymise; off = never export",
+        "default": "blocklist"
+      },
+      "redactToolNames": {
+        "type": "array",
+        "items": { "type": "string" },
+        "description": "Tool names to always redact regardless of toolNameExport mode (default: [])",
+        "default": []
+      },
+      "compressPayloads": {
+        "type": "boolean",
+        "description": "Gzip-compress batch payloads before sending (requires server-side Content-Encoding: gzip support, default: false)",
+        "default": false
       }
     },
     "additionalProperties": false

package/package.json

@@ -1,27 +1,40 @@
-{
-  "name": "agentmetrics-openclaw",
-  "version": "0.2.3",
-  "type": "module",
-  "description": "AgentMetrics observability plugin for OpenClaw agents",
-  "license": "MIT",
-  "homepage": "https://agentmetrics.dev",
-  "repository": {
-    "type": "git",
-    "url": "git+https://github.com/andalabx/agentmetrics-sdk.git"
-  },
-  "keywords": ["openclaw", "agentmetrics", "observability", "ai-agents", "monitoring"],
-  "peerDependencies": {
-    "openclaw": ">=2026.3.2"
-  },
-  "openclaw": {
-    "extensions": ["./index.ts"]
-  },
-  "files": [
-    "index.ts",
-    "openclaw.plugin.json",
-    "README.md"
-  ],
-  "engines": {
-    "node": ">=22"
-  }
-}
+{
+  "name": "agentmetrics-openclaw",
+  "version": "0.2.4",
+  "type": "module",
+  "description": "AgentMetrics observability plugin for OpenClaw agents",
+  "license": "MIT",
+  "homepage": "https://github.com/andalabx/agentmetrics",
+  "repository": {
+    "type": "git",
+    "url": "git+https://github.com/andalabx/agentmetrics.git"
+  },
+  "keywords": [
+    "openclaw",
+    "agentmetrics",
+    "observability",
+    "ai-agents",
+    "monitoring"
+  ],
+  "peerDependencies": {
+    "openclaw": ">=2026.3.2"
+  },
+  "peerDependenciesMeta": {
+    "openclaw": {
+      "optional": true
+    }
+  },
+  "openclaw": {
+    "extensions": [
+      "./index.ts"
+    ]
+  },
+  "files": [
+    "index.ts",
+    "openclaw.plugin.json",
+    "README.md"
+  ],
+  "engines": {
+    "node": ">=22"
+  }
+}