mcp-server-agentpay 1.0.4 → 1.1.0

package/index.js

@@ -10,6 +10,7 @@ const { version } = require("./package.json");
 // ── Config ──────────────────────────────────────────────────────────
 
 const GATEWAY_KEY = process.env.AGENTPAY_GATEWAY_KEY || "";
+const ADMIN_KEY = process.env.AGENTPAY_ADMIN_KEY || "";
 const BASE_URL = (process.env.AGENTPAY_URL || "https://agentpay.metaltorque.dev").replace(/\/$/, "");
 
 // ── HTTP helper ─────────────────────────────────────────────────────
@@ -76,6 +77,66 @@ function noKeyError() {
   };
 }
 
+function noAdminKeyError() {
+  return {
+    content: [{ type: "text", text: "Error: AGENTPAY_ADMIN_KEY environment variable is required for reliability endpoints.\n\nThis is the admin key configured on the AgentPay server." }],
+  };
+}
+
+function adminRequest(method, urlPath, body, timeout = 30_000) {
+  return new Promise((resolve, reject) => {
+    const fullUrl = `${BASE_URL}${urlPath}`;
+    const parsed = new URL(fullUrl);
+    const isHttps = parsed.protocol === "https:";
+
+    if (!isHttps && ADMIN_KEY) {
+      return reject(new Error("Refusing to send admin key over insecure HTTP. Use HTTPS."));
+    }
+
+    const mod = isHttps ? https : http;
+
+    const headers = {
+      "Content-Type": "application/json",
+      "User-Agent": `mcp-server-agentpay/${version}`,
+      "Authorization": `Bearer ${ADMIN_KEY}`,
+    };
+
+    const opts = {
+      hostname: parsed.hostname,
+      port: parsed.port || (isHttps ? 443 : 80),
+      path: parsed.pathname + parsed.search,
+      method,
+      headers,
+      timeout,
+    };
+
+    const req = mod.request(opts, (res) => {
+      let data = "";
+      let size = 0;
+      res.on("data", (c) => {
+        size += c.length;
+        if (size > MAX_RESPONSE_SIZE) { req.destroy(); return reject(new Error("Response too large")); }
+        data += c;
+      });
+      res.on("end", () => {
+        try {
+          const json = JSON.parse(data);
+          if (res.statusCode >= 400) return reject(new Error(json.error || `HTTP ${res.statusCode}`));
+          resolve(json);
+        } catch {
+          if (res.statusCode >= 400) return reject(new Error(`HTTP ${res.statusCode}: ${data.slice(0, 300)}`));
+          resolve(data);
+        }
+      });
+    });
+
+    req.on("error", reject);
+    req.on("timeout", () => { req.destroy(); reject(new Error("Request timed out")); });
+    if (body) req.write(JSON.stringify(body));
+    req.end();
+  });
+}
+
 // ── MCP Server ──────────────────────────────────────────────────────
 
 const server = new McpServer({
@@ -281,6 +342,151 @@ server.tool(
   }
 );
 
+// ── Tool: reliability_dashboard ────────────────────────────────────
+
+server.tool(
+  "reliability_dashboard",
+  "View the full reliability dashboard — circuit breaker states, health metrics (latency p50/p95/p99, success rates), and recent events for all tool backends. Requires admin key.",
+  {},
+  async () => {
+    if (!ADMIN_KEY) return noAdminKeyError();
+    try {
+      const result = await adminRequest("GET", "/admin/reliability");
+      const circuits = result.circuits || {};
+      const health = result.health || {};
+      const events = result.events || [];
+
+      let text = "=== Reliability Dashboard ===\n\n";
+
+      // Circuit states
+      text += "CIRCUITS:\n";
+      const toolIds = Object.keys(circuits);
+      if (toolIds.length === 0) {
+        text += "  No circuit data yet (no tool calls made since restart)\n";
+      } else {
+        for (const id of toolIds) {
+          const c = circuits[id];
+          text += `  ${id}: ${c.state} (failures: ${c.failureCount}, successes: ${c.successCount})\n`;
+        }
+      }
+
+      // Health metrics
+      text += "\nHEALTH:\n";
+      const healthIds = Object.keys(health);
+      if (healthIds.length === 0) {
+        text += "  No health data yet\n";
+      } else {
+        for (const id of healthIds) {
+          const h = health[id];
+          text += `  ${id}: ${h.successRate}% success (${h.requests.total} calls) — latency avg ${h.latency.avg}ms, p95 ${h.latency.p95}ms\n`;
+        }
+      }
+
+      // Recent events
+      if (events.length > 0) {
+        text += `\nRECENT EVENTS (last ${events.length}):\n`;
+        for (const e of events.slice(-10)) {
+          const status = e.success ? "OK" : "FAIL";
+          text += `  [${e.timestamp}] ${e.toolId}.${e.method} ${status} ${e.latencyMs}ms${e.error ? " — " + e.error : ""}\n`;
+        }
+        if (events.length > 10) text += `  ... and ${events.length - 10} more\n`;
+      }
+
+      return { content: [{ type: "text", text }] };
+    } catch (e) {
+      return { content: [{ type: "text", text: `Error: ${e.message}` }] };
+    }
+  }
+);
+
+// ── Tool: reliability_tool_detail ─────────────────────────────────
+
+server.tool(
+  "reliability_tool_detail",
+  "Get detailed reliability info for a specific tool — circuit breaker state, config (thresholds, timeouts), health metrics, and recent events.",
+  {
+    tool_id: z.string().describe("Tool ID (e.g. 'agent-audit', 'indexforge')"),
+  },
+  async ({ tool_id }) => {
+    if (!ADMIN_KEY) return noAdminKeyError();
+    try {
+      const result = await adminRequest("GET", `/admin/reliability/${encodeURIComponent(tool_id)}`);
+      const c = result.circuit || {};
+      const cfg = result.config || {};
+      const h = result.health || {};
+      const events = result.events || [];
+
+      let text = `=== ${tool_id} Reliability ===\n\n`;
+      text += `CIRCUIT: ${c.state} (failures: ${c.failureCount}, successes: ${c.successCount})\n`;
+      text += `CONFIG: threshold=${cfg.failureThreshold} failures, recovery=${cfg.recoveryTimeoutMs}ms, timeout=${cfg.requestTimeoutMs}ms\n`;
+      text += `HEALTH: ${h.successRate}% success (${h.requests?.total || 0} calls)\n`;
+      text += `LATENCY: avg=${h.latency?.avg || 0}ms, p50=${h.latency?.p50 || 0}ms, p95=${h.latency?.p95 || 0}ms, p99=${h.latency?.p99 || 0}ms (${h.latency?.samples || 0} samples)\n`;
+      if (h.lastCall) text += `LAST CALL: ${h.lastCall}\n`;
+
+      if (events.length > 0) {
+        text += `\nEVENTS (${events.length}):\n`;
+        for (const e of events.slice(-20)) {
+          const status = e.success ? "OK" : "FAIL";
+          text += `  [${e.timestamp}] ${e.method} ${status} ${e.latencyMs}ms${e.error ? " — " + e.error : ""}\n`;
+        }
+      }
+
+      return { content: [{ type: "text", text }] };
+    } catch (e) {
+      return { content: [{ type: "text", text: `Error: ${e.message}` }] };
+    }
+  }
+);
+
+// ── Tool: reliability_reset ───────────────────────────────────────
+
+server.tool(
+  "reliability_reset",
+  "Force-close a tripped circuit breaker for a tool, allowing calls to resume immediately. Use when a backend has recovered but the circuit hasn't automatically reset yet.",
+  {
+    tool_id: z.string().describe("Tool ID to reset (e.g. 'agent-audit', 'indexforge')"),
+  },
+  async ({ tool_id }) => {
+    if (!ADMIN_KEY) return noAdminKeyError();
+    try {
+      const result = await adminRequest("POST", `/admin/reliability/${encodeURIComponent(tool_id)}/reset`);
+      return { content: [{ type: "text", text: `Circuit reset for ${tool_id}: ${result.previousState} → ${result.currentState}` }] };
+    } catch (e) {
+      return { content: [{ type: "text", text: `Error: ${e.message}` }] };
+    }
+  }
+);
+
+// ── Tool: reliability_config ──────────────────────────────────────
+
+server.tool(
+  "reliability_config",
+  "Update reliability settings for a specific tool — failure threshold, recovery timeout, and request timeout. Changes take effect immediately.",
+  {
+    tool_id: z.string().describe("Tool ID to configure (e.g. 'agent-audit', 'indexforge')"),
+    failure_threshold: z.number().int().min(1).max(100).optional().describe("Number of consecutive failures before circuit opens (1-100)"),
+    recovery_timeout_ms: z.number().int().min(5000).max(600000).optional().describe("Milliseconds before OPEN circuit tries recovery (5000-600000)"),
+    request_timeout_ms: z.number().int().min(1000).max(600000).optional().describe("Per-request timeout in milliseconds (1000-600000)"),
+  },
+  async ({ tool_id, failure_threshold, recovery_timeout_ms, request_timeout_ms }) => {
+    if (!ADMIN_KEY) return noAdminKeyError();
+    const body = {};
+    if (failure_threshold !== undefined) body.failureThreshold = failure_threshold;
+    if (recovery_timeout_ms !== undefined) body.recoveryTimeoutMs = recovery_timeout_ms;
+    if (request_timeout_ms !== undefined) body.requestTimeoutMs = request_timeout_ms;
+    if (Object.keys(body).length === 0) {
+      return { content: [{ type: "text", text: "Error: Provide at least one setting to update (failure_threshold, recovery_timeout_ms, request_timeout_ms)" }] };
+    }
+    try {
+      const result = await adminRequest("POST", `/admin/reliability/${encodeURIComponent(tool_id)}/config`, body);
+      const cfg = result.config || {};
+      return { content: [{ type: "text", text: `Config updated for ${tool_id}:\n  failureThreshold: ${cfg.failureThreshold}\n  recoveryTimeoutMs: ${cfg.recoveryTimeoutMs}\n  requestTimeoutMs: ${cfg.requestTimeoutMs}` }] };
+    } catch (e) {
+      return { content: [{ type: "text", text: `Error: ${e.message}` }] };
+    }
+  }
+);
+
 // ── Start ───────────────────────────────────────────────────────────
 
 async function main() {

package/package.json

@@ -1,8 +1,8 @@
 {
   "name": "mcp-server-agentpay",
-  "version": "1.0.4",
+  "version": "1.1.0",
   "mcpName": "io.github.joepangallo/agent-pay",
-  "description": "MCP server for AgentPay — the payment gateway for autonomous AI agents. Lets agents discover, provision, and pay for MCP tool APIs with a single gateway key.",
+  "description": "MCP server for AgentPay — the payment gateway for autonomous AI agents. Discover, provision, and pay for MCP tool APIs. Includes reliability monitoring with circuit breakers and health metrics.",
   "bin": {
     "mcp-server-agentpay": "index.js"
   },

package/server.json

@@ -1,17 +1,17 @@
 {
   "$schema": "https://static.modelcontextprotocol.io/schemas/2025-12-11/server.schema.json",
   "name": "io.github.joepangallo/agent-pay",
-  "description": "Payment gateway for AI agents. Tool discovery, provisioning, and pay-per-call metering.",
+  "description": "Payment gateway for AI agents. Tool discovery, provisioning, pay-per-call metering, and reliability monitoring with circuit breakers.",
   "repository": {
     "url": "https://github.com/joepangallo/agent-pay",
     "source": "github"
   },
-  "version": "1.0.4",
+  "version": "1.1.0",
   "packages": [
     {
       "registryType": "npm",
       "identifier": "mcp-server-agentpay",
-      "version": "1.0.4",
+      "version": "1.1.0",
       "transport": {
         "type": "stdio"
       },
@@ -23,6 +23,13 @@
           "isSecret": true,
           "name": "AGENTPAY_GATEWAY_KEY"
         },
+        {
+          "description": "Admin key for reliability endpoints (circuit breakers, health metrics, config)",
+          "isRequired": false,
+          "format": "string",
+          "isSecret": true,
+          "name": "AGENTPAY_ADMIN_KEY"
+        },
         {
           "description": "AgentPay API URL",
           "isRequired": false,