MCP Datadog Playcourt

# MCP Datadog Server (SDK Version) MCP Server untuk Datadog menggunakan `@modelcontextprotocol/sdk` - implementasi resmi dari Model Context Protocol. ## Perbedaan dengan Versi Manual ### `mcp-server.js` (Manual Implementation) - Implementasi JSON-RPC manual - Handle request/response secara manual - Error handling custom ### `mcp-server-sdk.js` (SDK Implementation) ✨ **Recommended** - Menggunakan `@modelcontextprotocol/sdk` official - API yang lebih clean dan maintainable - Error handling otomatis dari SDK - Compatible dengan semua MCP clients ## Prerequisites ```bash npm install ``` Dependencies: - `@modelcontextprotocol/sdk`: ^1.0.1 - `axios`: ^1.6.0 - `dotenv`: ^16.0.3 ## Environment Variables Buat file `.env` di root folder: ```env DATADOG_API_KEY=your_api_key_here DATADOG_APP_KEY=your_app_key_here DATADOG_SITE=datadoghq.com ``` ## Running the Server ### Development ```bash npm run mcp-server-sdk ``` ### Integration dengan Claude Desktop Edit file konfigurasi Claude Desktop: - **MacOS**: `~/Library/Application Support/Claude/claude_desktop_config.json` - **Windows**: `%APPDATA%\Claude\claude_desktop_config.json` Tambahkan konfigurasi berikut: ```json { "mcpServers": { "datadog": { "command": "node", "args": [ "D:\\repo\\mcp-datadog\\datadog-mcp\\mcp-server-sdk.js" ], "env": { "DATADOG_API_KEY": "your_api_key", "DATADOG_APP_KEY": "your_app_key", "DATADOG_SITE": "datadoghq.com" } } } } ``` ## Available Tools ### 1. `list_services` List semua tools yang tersedia ### 2. `query_metrics` Query general purpose untuk metric Datadog apapun **Parameters:** - `query` (string): Datadog metric query - `from` (integer): Unix timestamp start (seconds) - `to` (integer): Unix timestamp end (seconds) **Example:** ```json { "query": "avg:system.cpu.user{*}", "from": 1700000000, "to": 1700003600 } ``` ### 3. `query_kubernetes_metrics` Query Kubernetes metrics **Parameters:** - `query` (string): Datadog K8s metric query - `from` (integer): Unix timestamp start - `to` (integer): Unix timestamp end - `cluster` (string, optional): Cluster name ### 4. `query_deployment_pods` Query pod count dan metrics untuk deployment spesifik **Parameters:** - `namespace` (string): K8s namespace - `deployment` (string): Deployment name - `from` (integer): Unix timestamp start - `to` (integer): Unix timestamp end - `cluster` (string, optional): Cluster name - `apm_service_name` (string, optional): APM service name **Returns:** - pods count - CPU usage & limits - Memory usage & limits - Network in/out - Response time (APM) ### 5. `query_apm_metrics` Query APM metrics untuk service **Parameters:** - `service` (string): Service name - `from` (integer): Unix timestamp start - `to` (integer): Unix timestamp end - `metric` (string, optional): Metric name - `aggregator` (string, optional): avg|max|min|sum ### 6. `query_host_metrics` Query host/infrastructure metrics **Parameters:** - `hostname` (string): Hostname - `from` (integer): Unix timestamp start - `to` (integer): Unix timestamp end - `metric` (string, optional): Metric name - `aggregator` (string, optional): avg|max|min|sum ### 7. `query_database_metrics` Query database monitoring metrics **Parameters:** - `db_host` (string): Database host - `from` (integer): Unix timestamp start - `to` (integer): Unix timestamp end - `db_type` (string, optional): postgresql|mysql - `metric` (string, optional): Metric name - `aggregator` (string, optional): avg|max|min|sum ### 8. `list_hosts` List semua monitored hosts **Parameters:** - `from` (integer): Unix timestamp start - `to` (integer): Unix timestamp end ### 9. `list_databases` List semua monitored databases **Parameters:** - `from` (integer): Unix timestamp start - `to` (integer): Unix timestamp end - `db_type` (string, optional): postgresql|mysql ### 10. `list_all_hosts` List semua hosts dengan informasi lengkap (CPU, memory, OS, kernel, dll) **Parameters:** None ### 11. `get_host_detail` Get detail lengkap untuk host tertentu **Parameters:** - `hostname` (string): Hostname ### 12. `get_namespace_metrics` Get total resource usage untuk K8s namespace **Parameters:** - `namespace` (string): K8s namespace - `from` (integer): Unix timestamp start - `to` (integer): Unix timestamp end - `cluster` (string, optional): Cluster name ## Testing ### Manual Test ```bash # Initialize echo '{"jsonrpc":"2.0","id":1,"method":"initialize","params":{"protocolVersion":"2024-11-05","capabilities":{},"clientInfo":{"name":"test","version":"1.0"}}}' | node mcp-server-sdk.js # List tools echo '{"jsonrpc":"2.0","id":2,"method":"tools/list","params":{}}' | node mcp-server-sdk.js # Call tool echo '{"jsonrpc":"2.0","id":3,"method":"tools/call","params":{"name":"list_all_hosts","arguments":{}}}' | node mcp-server-sdk.js ``` ## Architecture ``` mcp-server-sdk.js ├── @modelcontextprotocol/sdk │ ├── Server │ ├── StdioServerTransport │ └── Request Schemas ├── Datadog API Client (axios) ├── CLI Query Executor (npm run query) └── Tool Handlers ├── listServices() ├── queryMetrics() ├── queryKubernetesMetrics() ├── queryDeploymentPods() ├── queryAPMMetrics() ├── queryHostMetrics() ├── queryDatabaseMetrics() ├── listHosts() ├── listDatabases() ├── listAllHosts() ├── getHostDetail() └── getNamespaceMetrics() ``` ## Migration dari Manual ke SDK Jika saat ini menggunakan `mcp-server.js`, untuk migrasi ke SDK version: 1. Update Claude Desktop config untuk menggunakan `mcp-server-sdk.js` 2. Restart Claude Desktop 3. Test dengan query sederhana Tidak ada perubahan di API atau parameter - 100% backward compatible! ## Advantages of SDK Version ✅ **Maintainable**: Kode lebih clean dan terstruktur ✅ **Reliable**: Error handling otomatis dari SDK ✅ **Future-proof**: Update SDK untuk fitur baru MCP ✅ **Standard**: Mengikuti best practices MCP ✅ **Type-safe**: Better TypeScript support (jika migrasi ke TS) ## Next Steps Untuk menambahkan tool baru: 1. Tambahkan tool definition di `ListToolsRequestSchema` handler 2. Tambahkan case di `CallToolRequestSchema` handler 3. Implementasikan function handler 4. Test dengan MCP Inspector atau Claude Desktop ## Troubleshooting ### Error: Cannot find module '@modelcontextprotocol/sdk' ```bash npm install ``` ### Error: DATADOG_API_KEY not set Pastikan file `.env` sudah dibuat dengan API keys yang valid ### Server tidak respond Check console error dengan: ```bash node mcp-server-sdk.js 2>&1 | tee debug.log ``` ## Support Untuk pertanyaan atau issue, check: - [MCP Documentation](https://modelcontextprotocol.io) - [Datadog API Docs](https://docs.datadoghq.com/api/)