MCP Datadog Playcourt

#!/usr/bin/env node /** * MCP Datadog Server (using @modelcontextprotocol/sdk) * * Runs as an MCP (Model Context Protocol) server using the official SDK. * Compatible with Claude Desktop, VS Code, and other MCP clients. */ import { Server } from "@modelcontextprotocol/sdk/server/index.js"; import { StdioServerTransport } from "@modelcontextprotocol/sdk/server/stdio.js"; import { CallToolRequestSchema, ListToolsRequestSchema, } from "@modelcontextprotocol/sdk/types.js"; import dotenv from 'dotenv'; import axios from 'axios'; import { execSync } from 'child_process'; import { fileURLToPath } from 'url'; import path from 'path'; // Load environment variables dotenv.config(); const API_KEY = process.env.DATADOG_API_KEY; const APP_KEY = process.env.DATADOG_APP_KEY; const SITE = process.env.DATADOG_SITE || 'datadoghq.com'; if (!API_KEY || !APP_KEY) { console.error('Error: DATADOG_API_KEY and DATADOG_APP_KEY must be set in .env file'); process.exit(1); } // Datadog API client const datadogClient = axios.create({ baseURL: `https://api.${SITE}`, headers: { 'DD-API-KEY': API_KEY, 'DD-APPLICATION-KEY': APP_KEY, 'Content-Type': 'application/json' } }); const __filename = fileURLToPath(import.meta.url); const __dirname = path.dirname(__filename); // Helper function to execute npm scripts function executeCliQuery(query, from, to) { try { // Validate timestamps are in seconds (not milliseconds) const now = Math.floor(Date.now() / 1000); const maxFuture = now + (24 * 3600); // Allow up to 24 hours in future if (from <= 0 || to <= 0) { throw new Error(`Invalid timestamp range: from=${from}, to=${to}. Timestamps must be positive.`); } if (from > 2000000000 || to > 2000000000) { throw new Error(`Invalid timestamp: from=${from}, to=${to}. Looks like milliseconds. Use seconds instead.`); } if (from > maxFuture || to > maxFuture) { const fromDate = new Date(from * 1000).toISOString(); const toDate = new Date(to * 1000).toISOString(); throw new Error(`Timestamp too far in future: from=${fromDate}, to=${toDate}. Current time=${new Date(now * 1000).toISOString()}`); } const cmd = `npm run query -- "${query}" ${from} ${to}`; const output = execSync(cmd, { cwd: __dirname, encoding: 'utf-8' }); return output; } catch (error) { throw new Error(`Query execution failed: ${error.message}`); } } // Create MCP Server const server = new Server( { name: "datadog-mcp", version: "2.0.0" }, { capabilities: { tools: {} } } ); // Advertise available tools server.setRequestHandler(ListToolsRequestSchema, async () => { return { tools: [ { name: "list_services", description: "List all available tools/services in this MCP server", inputSchema: { type: "object", properties: {}, required: [] } }, { name: "query_metrics", description: "General purpose query for any Datadog metric. Use this when other specific tools don't match your needs.", inputSchema: { type: "object", properties: { query: { type: "string", description: "Datadog metric query string (e.g., avg:system.cpu.user{*}, sum:kubernetes.pods.running{kube_namespace:default})" }, from: { type: "integer", description: "Start time in Unix timestamp (seconds)" }, to: { type: "integer", description: "End time in Unix timestamp (seconds)" } }, required: ["query", "from", "to"] } }, { name: "query_kubernetes_metrics", description: "Query Kubernetes metrics from Datadog", inputSchema: { type: "object", properties: { query: { type: "string", description: "Datadog metric query (e.g., avg:kubernetes.cpu.usage.total{kube_namespace:pijarsekolah-prod})" }, from: { type: "integer", description: "Start time in Unix timestamp (seconds)" }, to: { type: "integer", description: "End time in Unix timestamp (seconds)" }, cluster: { type: "string", description: "Kubernetes cluster name (optional)" } }, required: ["query", "from", "to"] } }, { name: "query_deployment_pods", description: "Query pod count for a specific Kubernetes deployment", inputSchema: { type: "object", properties: { namespace: { type: "string", description: "Kubernetes namespace (e.g., pijarsekolah-prod)" }, deployment: { type: "string", description: "Deployment name (e.g., pijarsekolah-cbt-exam-api)" }, from: { type: "integer", description: "Start time in Unix timestamp (seconds)" }, to: { type: "integer", description: "End time in Unix timestamp (seconds)" }, cluster: { type: "string", description: "Kubernetes cluster name (optional)" }, apm_service_name: { type: "string", description: "APM service name for response time metric (optional, defaults to deployment name)" } }, required: ["namespace", "deployment", "from", "to"] } }, { name: "query_apm_metrics", description: "Query APM metrics for a specific service", inputSchema: { type: "object", properties: { service: { type: "string", description: "Service name" }, from: { type: "integer", description: "Start time in Unix timestamp (seconds)" }, to: { type: "integer", description: "End time in Unix timestamp (seconds)" }, metric: { type: "string", description: "Metric name (e.g., trace.web.request.duration)" }, aggregator: { type: "string", enum: ["avg", "max", "min", "sum"], description: "Aggregation function" } }, required: ["service", "from", "to"] } }, { name: "query_host_metrics", description: "Query host/infrastructure metrics", inputSchema: { type: "object", properties: { hostname: { type: "string", description: "Hostname to query" }, from: { type: "integer", description: "Start time in Unix timestamp (seconds)" }, to: { type: "integer", description: "End time in Unix timestamp (seconds)" }, metric: { type: "string", description: "Metric name (e.g., system.cpu.user)" }, aggregator: { type: "string", enum: ["avg", "max", "min", "sum"], description: "Aggregation function" } }, required: ["hostname", "from", "to"] } }, { name: "query_database_metrics", description: "Query database monitoring (DBM) metrics", inputSchema: { type: "object", properties: { db_host: { type: "string", description: "Database host name" }, from: { type: "integer", description: "Start time in Unix timestamp (seconds)" }, to: { type: "integer", description: "End time in Unix timestamp (seconds)" }, db_type: { type: "string", enum: ["postgresql", "mysql"], description: "Database type" }, metric: { type: "string", description: "Metric name" }, aggregator: { type: "string", enum: ["avg", "max", "min", "sum"], description: "Aggregation function" } }, required: ["db_host", "from", "to"] } }, { name: "list_hosts", description: "List all monitored hosts", inputSchema: { type: "object", properties: { from: { type: "integer", description: "Start time in Unix timestamp (seconds)" }, to: { type: "integer", description: "End time in Unix timestamp (seconds)" } }, required: ["from", "to"] } }, { name: "list_databases", description: "List all monitored databases", inputSchema: { type: "object", properties: { from: { type: "integer", description: "Start time in Unix timestamp (seconds)" }, to: { type: "integer", description: "End time in Unix timestamp (seconds)" }, db_type: { type: "string", enum: ["postgresql", "mysql"], description: "Database type" } }, required: ["from", "to"] } }, { name: "list_all_hosts", description: "List all monitored hosts with complete system info (hostname, CPU, memory, disk, OS, kernel, applications, etc)", inputSchema: { type: "object", properties: {}, required: [] } }, { name: "get_host_detail", description: "Get detailed information for a specific host by hostname", inputSchema: { type: "object", properties: { hostname: { type: "string", description: "Hostname to get details for" } }, required: ["hostname"] } }, { name: "get_namespace_metrics", description: "Get total resource usage and limits for a Kubernetes namespace", inputSchema: { type: "object", properties: { namespace: { type: "string", description: "Kubernetes namespace (e.g., pijarsekolah-prod)" }, from: { type: "integer", description: "Start time in Unix timestamp (seconds)" }, to: { type: "integer", description: "End time in Unix timestamp (seconds)" }, cluster: { type: "string", description: "Kubernetes cluster name (optional)" } }, required: ["namespace", "from", "to"] } } ] }; }); // Handle tool calls server.setRequestHandler(CallToolRequestSchema, async (request) => { const tool = request.params.name; const args = (request.params.arguments ?? {}); console.error(`[DEBUG] Tool called: ${tool}`); console.error(`[DEBUG] Arguments:`, JSON.stringify(args, null, 2)); try { switch (tool) { case 'list_services': return await listServices(); case 'query_metrics': return await queryMetrics(args); case 'query_kubernetes_metrics': return await queryKubernetesMetrics(args); case 'query_deployment_pods': return await queryDeploymentPods(args); case 'query_apm_metrics': return await queryAPMMetrics(args); case 'query_host_metrics': return await queryHostMetrics(args); case 'query_database_metrics': return await queryDatabaseMetrics(args); case 'list_hosts': return await listHosts(args); case 'list_databases': return await listDatabases(args); case 'list_all_hosts': return await listAllHosts(); case 'get_host_detail': return await getHostDetail(args); case 'get_namespace_metrics': return await getNamespaceMetrics(args); default: throw new Error(`Unknown tool: ${tool}`); } } catch (error) { console.error(`[ERROR] Tool execution failed:`, error.message); console.error(`[ERROR] Stack:`, error.stack); return { content: [ { type: "text", text: `Error executing tool: ${error.message}` } ], isError: true }; } }); // Tool implementations async function listServices() { const services = [ { name: "list_services", description: "List all available tools/services" }, { name: "query_metrics", description: "General purpose query for any Datadog metric" }, { name: "query_kubernetes_metrics", description: "Query Kubernetes metrics" }, { name: "query_deployment_pods", description: "Query pod count for a deployment" }, { name: "query_apm_metrics", description: "Query APM metrics" }, { name: "query_host_metrics", description: "Query host/infrastructure metrics" }, { name: "query_database_metrics", description: "Query database monitoring metrics" }, { name: "list_hosts", description: "List all monitored hosts" }, { name: "list_databases", description: "List all monitored databases" }, { name: "list_all_hosts", description: "List all hosts with complete system info" }, { name: "get_host_detail", description: "Get detailed host information" }, { name: "get_namespace_metrics", description: "Get namespace resource usage" } ]; return { content: [ { type: "text", text: JSON.stringify({ total_services: services.length, services }, null, 2) } ] }; } async function queryMetrics(args) { const from = parseInt(args.from); const to = parseInt(args.to); const now = Math.floor(Date.now() / 1000); const maxFuture = now + (24 * 3600); console.error(`[DEBUG] queryMetrics - from: ${from}, to: ${to}, now: ${now}`); console.error(`[DEBUG] queryMetrics - query: ${args.query}`); if (from <= 0 || to <= 0 || from > 2000000000 || to > 2000000000) { console.error(`[ERROR] Invalid timestamp detected`); return { content: [ { type: "text", text: `Error: Invalid timestamp. from=${from}, to=${to}. Use Unix timestamp in seconds (not milliseconds).` } ] }; } if (from > maxFuture || to > maxFuture) { console.error(`[ERROR] Timestamp too far in future`); return { content: [ { type: "text", text: `Error: Timestamp in future. from=${new Date(from * 1000).toISOString()}, to=${new Date(to * 1000).toISOString()}. Current time=${new Date(now * 1000).toISOString()}` } ] }; } const query = args.query; const output = executeCliQuery(query, from, to); console.error(`[DEBUG] Query result length: ${output.length} chars`); return { content: [{ type: "text", text: output }] }; } async function queryKubernetesMetrics(args) { const from = parseInt(args.from); const to = parseInt(args.to); const now = Math.floor(Date.now() / 1000); const maxFuture = now + (24 * 3600); console.error(`[DEBUG] queryKubernetesMetrics - from: ${from}, to: ${to}, now: ${now}`); console.error(`[DEBUG] queryKubernetesMetrics - original query: ${args.query}`); if (from <= 0 || to <= 0 || from > 2000000000 || to > 2000000000) { console.error(`[ERROR] Invalid timestamp detected`); return { content: [ { type: "text", text: `Error: Invalid timestamp. from=${from}, to=${to}. Use Unix timestamp in seconds (not milliseconds).` } ] }; } if (from > maxFuture || to > maxFuture) { console.error(`[ERROR] Timestamp too far in future`); return { content: [ { type: "text", text: `Error: Timestamp in future. from=${new Date(from * 1000).toISOString()}, to=${new Date(to * 1000).toISOString()}. Current=${new Date(now * 1000).toISOString()}` } ] }; } let query = args.query; // Add cluster filter if provided if (args.cluster) { if (query.includes('{')) { query = query.replace('{', `{kube_cluster_name:${args.cluster},`); } else { query += `{kube_cluster_name:${args.cluster}}`; } } console.error(`[DEBUG] queryKubernetesMetrics - final query: ${query}`); const output = executeCliQuery(query, from, to); console.error(`[DEBUG] Query result length: ${output.length} chars`); return { content: [{ type: "text", text: output }] }; } async function queryDeploymentPods(args) { const namespace = args.namespace; const deployment = args.deployment; const cluster = args.cluster || null; const apmServiceName = args.apm_service_name || null; const from = args.from; const to = args.to; console.error(`[DEBUG] queryDeploymentPods - namespace: ${namespace}, deployment: ${deployment}`); console.error(`[DEBUG] queryDeploymentPods - from: ${from}, to: ${to}`); console.error(`[DEBUG] queryDeploymentPods - cluster: ${cluster}, apmServiceName: ${apmServiceName}`); const clusterFilter = cluster ? `, kube_cluster_name:${cluster}` : ''; const podQuery = `sum:kubernetes.pods.running{kube_namespace:${namespace}, kube_deployment:${deployment}${clusterFilter}}`; const cpuQuery = `avg:kubernetes.cpu.usage.total{kube_namespace:${namespace}, kube_deployment:${deployment}${clusterFilter}}`; const memoryQuery = `avg:kubernetes.memory.usage{kube_namespace:${namespace}, kube_deployment:${deployment}${clusterFilter}}`; const cpuLimitQuery = `avg:kubernetes.cpu.limits{kube_namespace:${namespace}, kube_deployment:${deployment}${clusterFilter}}`; const memoryLimitQuery = `avg:kubernetes.memory.limits{kube_namespace:${namespace}, kube_deployment:${deployment}${clusterFilter}}`; const networkInQuery = `sum:kubernetes.network.rx_bytes{kube_namespace:${namespace}, kube_deployment:${deployment}${clusterFilter}}`; const networkOutQuery = `sum:kubernetes.network.tx_bytes{kube_namespace:${namespace}, kube_deployment:${deployment}${clusterFilter}}`; let envName = 'production'; if (namespace.includes('-prod')) { envName = 'production'; } else if (namespace.includes('-stage')) { envName = 'staging'; } const serviceName = apmServiceName || deployment; const responseTimeMetric = 'trace.http.request'; let result = { cluster: cluster, namespace: namespace, deployment: deployment, pods: 'N/A', cpu_usage: 'N/A', cpu_limit: 'N/A', memory_usage: 'N/A', memory_limit: 'N/A', network_in: 'N/A', network_out: 'N/A', response_time: 'N/A' }; try { result.pods = executeCliQuery(podQuery, from, to).trim(); } catch (e) { result.pods = `Error: ${e.message}`; } try { result.cpu_usage = executeCliQuery(cpuQuery, from, to).trim(); } catch (e) { result.cpu_usage = `Error: ${e.message}`; } try { result.cpu_limit = executeCliQuery(cpuLimitQuery, from, to).trim(); } catch (e) { result.cpu_limit = `Error: ${e.message}`; } try { result.memory_usage = executeCliQuery(memoryQuery, from, to).trim(); } catch (e) { result.memory_usage = `Error: ${e.message}`; } try { result.memory_limit = executeCliQuery(memoryLimitQuery, from, to).trim(); } catch (e) { result.memory_limit = `Error: ${e.message}`; } try { result.network_in = executeCliQuery(networkInQuery, from, to).trim(); } catch (e) { result.network_in = `Error: ${e.message}`; } try { result.network_out = executeCliQuery(networkOutQuery, from, to).trim(); } catch (e) { result.network_out = `Error: ${e.message}`; } try { const query = `max:${responseTimeMetric}{service:${serviceName}, env:${envName}${cluster ? `, kube_cluster_name:${cluster}` : ''}}`; const responseTimeOutput = executeCliQuery(query, from, to); const trimmed = responseTimeOutput.trim(); if (trimmed && !trimmed.includes('No data found')) { result.response_time = trimmed; } else { result.response_time = 'No data'; } } catch (e) { result.response_time = `Error: ${e.message}`; } return { content: [{ type: "text", text: JSON.stringify(result, null, 2) }] }; } async function queryAPMMetrics(args) { const metric = args.metric || 'avg:trace.web.request.duration'; const service = args.service || '*'; const query = `${metric}{service:${service}}`; const output = executeCliQuery(query, args.from, args.to); return { content: [{ type: "text", text: output }] }; } async function queryHostMetrics(args) { const metric = args.metric || 'avg:system.cpu.user'; const hostname = args.hostname || '*'; const query = `${metric}{host:${hostname}}`; const output = executeCliQuery(query, args.from, args.to); return { content: [{ type: "text", text: output }] }; } async function queryDatabaseMetrics(args) { const dbType = args.db_type || 'postgresql'; const metric = args.metric || `avg:${dbType}.queries.active`; const dbHost = args.db_host || '*'; const query = `${metric}{dbhost:${dbHost}}`; const output = executeCliQuery(query, args.from, args.to); return { content: [{ type: "text", text: output }] }; } async function listHosts(args) { const query = `avg:system.cpu.user{*} by {host}`; const output = executeCliQuery(query, args.from, args.to); return { content: [{ type: "text", text: output }] }; } async function listDatabases(args) { const dbType = args.db_type || 'postgresql'; const query = `avg:${dbType}.queries.active{*} by {dbhost}`; const output = executeCliQuery(query, args.from, args.to); return { content: [{ type: "text", text: output }] }; } async function listAllHosts() { const response = await datadogClient.get('/api/v1/hosts'); const hosts = response.data.host_list || []; const hostInfo = hosts.map(host => ({ hostname: host.name, status: host.up ? 'up' : 'down', last_reported: host.last_reported_time ? new Date(host.last_reported_time * 1000).toISOString() : 'unknown', os: host.meta?.platform || 'unknown', processor: host.meta?.processor || 'unknown', kernel: host.meta?.kernel_release || 'unknown', agent_version: host.meta?.agent_version || 'unknown', applications: (host.apps || []).slice(0, 8).join(', '), tags: (host.tags || []).join(', ') || 'none', cpu_cores: host.meta?.cpuCores || 'unknown', total_memory: host.meta?.memory?.total ? (parseInt(host.meta.memory.total.replace(/[^0-9]/g, '')) / 1024).toFixed(2) + ' GB' : 'unknown', cpu_usage: host.metrics?.cpu ? host.metrics.cpu.toFixed(2) + '%' : 'N/A', memory_usage: host.metrics?.iowait ? host.metrics.iowait.toFixed(2) + '%' : 'N/A', host_id: host.meta?.host_id || 'unknown' })); hostInfo.sort((a, b) => a.hostname.localeCompare(b.hostname)); return { content: [{ type: "text", text: JSON.stringify(hostInfo, null, 2) }] }; } async function getHostDetail(args) { const response = await datadogClient.get('/api/v1/hosts'); const hosts = response.data.host_list || []; const host = hosts.find(h => h.name === args.hostname || h.name.toLowerCase() === args.hostname.toLowerCase()); if (!host) { return { content: [{ type: "text", text: `Host '${args.hostname}' not found` }] }; } const hostDetail = { hostname: host.name, status: host.up ? 'up' : 'down', last_reported: host.last_reported_time ? new Date(host.last_reported_time * 1000).toISOString() : 'unknown', os: host.meta?.platform || 'unknown', processor: host.meta?.processor || 'unknown', kernel: host.meta?.kernel_release || 'unknown', agent_version: host.meta?.agent_version || 'unknown', applications: (host.apps || []).join(', ') || 'none', tags: (host.tags || []).join(', ') || 'none', cpu_cores: host.meta?.cpuCores || 'unknown', total_memory: host.meta?.memory?.total ? (parseInt(host.meta.memory.total.replace(/[^0-9]/g, '')) / 1024).toFixed(2) + ' GB' : 'unknown', cpu_usage: host.metrics?.cpu ? host.metrics.cpu.toFixed(2) + '%' : 'N/A', memory_usage: host.metrics?.iowait ? host.metrics.iowait.toFixed(2) + '%' : 'N/A', host_id: host.meta?.host_id || 'unknown', socket_fqdn: host.meta?.socket_fqdn || 'unknown', socket_hostname: host.meta?.socket_hostname || 'unknown' }; return { content: [{ type: "text", text: JSON.stringify(hostDetail, null, 2) }] }; } async function getNamespaceMetrics(args) { const namespace = args.namespace; const cluster = args.cluster || null; const from = args.from; const to = args.to; const clusterFilter = cluster ? `, kube_cluster_name:${cluster}` : ''; const podQuery = `sum:kubernetes.pods.running{kube_namespace:${namespace}${clusterFilter}}`; const cpuUsageQuery = `sum:kubernetes.cpu.usage.total{kube_namespace:${namespace}${clusterFilter}}`; const cpuLimitQuery = `sum:kubernetes.cpu.limits{kube_namespace:${namespace}${clusterFilter}}`; const memoryUsageQuery = `sum:kubernetes.memory.usage{kube_namespace:${namespace}${clusterFilter}}`; const memoryLimitQuery = `sum:kubernetes.memory.limits{kube_namespace:${namespace}${clusterFilter}}`; const networkInQuery = `sum:kubernetes.network.rx_bytes{kube_namespace:${namespace}${clusterFilter}}`; const networkOutQuery = `sum:kubernetes.network.tx_bytes{kube_namespace:${namespace}${clusterFilter}}`; let result = { cluster: cluster, namespace: namespace, total_pods: 'N/A', cpu_usage_total: 'N/A', cpu_limit_total: 'N/A', memory_usage_total: 'N/A', memory_limit_total: 'N/A', network_in_total: 'N/A', network_out_total: 'N/A' }; try { result.total_pods = executeCliQuery(podQuery, from, to).trim(); } catch (e) { result.total_pods = `Error: ${e.message}`; } try { result.cpu_usage_total = executeCliQuery(cpuUsageQuery, from, to).trim(); } catch (e) { result.cpu_usage_total = `Error: ${e.message}`; } try { result.cpu_limit_total = executeCliQuery(cpuLimitQuery, from, to).trim(); } catch (e) { result.cpu_limit_total = `Error: ${e.message}`; } try { result.memory_usage_total = executeCliQuery(memoryUsageQuery, from, to).trim(); } catch (e) { result.memory_usage_total = `Error: ${e.message}`; } try { result.memory_limit_total = executeCliQuery(memoryLimitQuery, from, to).trim(); } catch (e) { result.memory_limit_total = `Error: ${e.message}`; } try { result.network_in_total = executeCliQuery(networkInQuery, from, to).trim(); } catch (e) { result.network_in_total = `Error: ${e.message}`; } try { result.network_out_total = executeCliQuery(networkOutQuery, from, to).trim(); } catch (e) { result.network_out_total = `Error: ${e.message}`; } return { content: [{ type: "text", text: JSON.stringify(result, null, 2) }] }; } // Run the server async function run() { const transport = new StdioServerTransport(); await server.connect(transport); console.error("MCP Datadog Server running on stdio"); } run().catch((err) => { console.error("Fatal error:", err); process.exit(1); });