Google Cloud Docs MCP Server

Instructions

Implementation Reference

• src/index.ts:912-1016 (handler)

  The handler function that executes the get_api_reference tool. It maps services to API reference paths, fetches documentation using fetchGoogleCloudDoc, and returns structured JSON with API details, resources, and content.
  async function getApiReference( service: string, resource?: string ): Promise<string> { const serviceLower = service.toLowerCase(); const product = GOOGLE_CLOUD_PRODUCTS[serviceLower]; if (!product) { return JSON.stringify({ error: `Unknown service: ${service}`, availableServices: Object.keys(GOOGLE_CLOUD_PRODUCTS), suggestion: 'Use "list_google_cloud_products" to see all available services.', }); } // API reference paths for common services const apiPaths: Record<string, { rest: string; resources: string[] }> = { compute: { rest: "compute/docs/reference/rest/v1", resources: [ "instances", "disks", "networks", "firewalls", "images", "machineTypes", ], }, storage: { rest: "storage/docs/json_api/v1", resources: ["buckets", "objects", "notifications"], }, bigquery: { rest: "bigquery/docs/reference/rest", resources: ["datasets", "tables", "jobs", "routines"], }, pubsub: { rest: "pubsub/docs/reference/rest", resources: ["topics", "subscriptions", "snapshots"], }, sql: { rest: "sql/docs/mysql/admin-api/rest/v1", resources: ["instances", "databases", "users", "backupRuns"], }, kubernetes: { rest: "kubernetes-engine/docs/reference/rest", resources: ["clusters", "nodePools", "operations"], }, functions: { rest: "functions/docs/reference/rest/v2", resources: ["functions", "operations", "locations"], }, run: { rest: "run/docs/reference/rest", resources: ["services", "configurations", "routes", "revisions"], }, iam: { rest: "iam/docs/reference/rest", resources: ["roles", "serviceAccounts", "policies"], }, }; const apiInfo = apiPaths[serviceLower]; if (!apiInfo) { return JSON.stringify({ service: product.name, docsUrl: `https://cloud.google.com/${product.docsPath}`, apiReference: `https://cloud.google.com/${product.docsPath}/reference`, note: "API reference path not pre-configured. Try fetching the docs URL directly.", }); } const apiUrl = resource ? `https://cloud.google.com/${apiInfo.rest}/${resource}` : `https://cloud.google.com/${apiInfo.rest}`; // Fetch the actual API reference page try { const docContent = await fetchGoogleCloudDoc(apiInfo.rest); const parsedContent = JSON.parse(docContent); return JSON.stringify({ service: product.name, description: product.description, apiReferenceUrl: apiUrl, availableResources: apiInfo.resources, selectedResource: resource || "overview", documentation: parsedContent, usage: resource ? `Viewing API reference for ${resource}` : `Use "get_api_reference" with a resource parameter to get specific resource documentation. Available: ${apiInfo.resources.join(", ")}`, }); } catch { return JSON.stringify({ service: product.name, description: product.description, apiReferenceUrl: apiUrl, availableResources: apiInfo.resources, selectedResource: resource || "overview", fetchCommand: `Use fetch_google_cloud_doc with path: "${resource ? `${apiInfo.rest}/${resource}` : apiInfo.rest}"`, }); } }
• src/index.ts:156-207 (schema)

  The tool registration including name, detailed description, and input schema definition (service required, resource optional) used by the MCP server for listing and validating tool calls.
  name: "get_api_reference", description: `Get REST API reference documentation for a specific Google Cloud service. **WHEN TO USE**: Use this tool when: - User needs API endpoints, methods, or parameters - User is developing integrations with GCP APIs - User asks about REST API for a specific GCP service - User needs to know available API resources for a service **INPUT**: - service (required): GCP service name (compute, storage, bigquery, pubsub, sql, kubernetes, functions, run, iam) - resource (optional): Specific API resource (instances, buckets, datasets, topics, etc.) **SUPPORTED SERVICES & RESOURCES**: - compute: instances, disks, networks, firewalls, images, machineTypes - storage: buckets, objects, notifications - bigquery: datasets, tables, jobs, routines - pubsub: topics, subscriptions, snapshots - sql: instances, databases, users, backupRuns - kubernetes: clusters, nodePools, operations - functions: functions, operations, locations - run: services, configurations, routes, revisions - iam: roles, serviceAccounts, policies **OUTPUT**: Returns JSON with: - service: Service name - description: Service description - apiReferenceUrl: Full URL to API reference - availableResources: List of available resources for this service - documentation: Actual API documentation content (if available) **EXAMPLE USAGE**: - Get Compute Engine API overview: service="compute" - Get Storage buckets API: service="storage", resource="buckets" - Get BigQuery datasets API: service="bigquery", resource="datasets"`, inputSchema: { type: "object" as const, properties: { service: { type: "string", description: "The Google Cloud service name (e.g., 'compute', 'storage', 'bigquery', 'pubsub', 'sql', 'kubernetes', 'functions', 'run', 'iam')", }, resource: { type: "string", description: "Optional: Specific API resource (e.g., 'instances', 'buckets', 'datasets', 'topics')", }, }, required: ["service"], }, },
• src/index.ts:1065-1074 (registration)

  The switch case that handles incoming 'get_api_reference' tool calls, extracts parameters, invokes the handler, and returns the result as MCP content.
  case "get_api_reference": { const { service, resource } = args as { service: string; resource?: string; }; const result = await getApiReference(service, resource); return { content: [{ type: "text", text: result }], }; }