Cloud Run MCP Server is an MCP server in the official Google Cloud Platform repository, designed to enable MCP-compatible AI agents to deploy applications to Google Cloud Run. It supports deployment from Gemini CLI, other AI-powered CLI agents, AI-powered IDEs, AI assistant applications, and agent SDKs such as the Google Gen AI SDK or Agent Development Kit.
Key capabilities include the following tools: deploy-file-contents (deploy directly to Cloud Run by providing file contents), list-services (list Cloud Run services in a specified project and region), get-service (get detailed information about a specific Cloud Run service), get-service-log (get logs and error messages for a specific Cloud Run service), deploy-local-folder (deploy a local folder to a Cloud Run service, available only when running locally), list-projects (list available GCP projects, available only when running locally), create-project (create a new GCP project and attach it to the first available billing account, optionally specifying a project ID, available only when running locally). It also provides two natural language prompts: deploy (deploy the current working directory to Cloud Run; if no service name is provided, use the DEFAULT_SERVICE_NAME environment variable or the current working directory name) and logs (get logs for a Cloud Run service; if no service name is provided, use the DEFAULT_SERVICE_NAME environment variable or the current working directory name).
Usage includes: installing as a Gemini CLI extension (run gemini extensions install https://github.com/GoogleCloudPlatform/cloud-run-mcp，然后使用 gcloud auth login and gcloud auth application-default login to log in); using it in an MCP client, which can be configured as a local MCP server (run via Node.js or Docker, using local Google Cloud credentials) or a remote MCP server (deploy the MCP server to Cloud Run and connect locally through IAM authentication, or use OAuth authentication). Example local MCP server configuration: "cloud-run": { "command": "npx", "args": ["-y", "@google-cloud/cloud-run-mcp"] }, and you can add the environment variables GOOGLE_CLOUD_PROJECT, GOOGLE_CLOUD_REGION, DEFAULT_SERVICE_NAME. The remote MCP server must first be deployed: gcloud run deploy cloud-run-mcp --image us-docker.pkg.dev/cloudrun/container/mcp --no-allow-unauthenticated, then create a local proxy through gcloud run services proxy, and configure "cloud-run": { "url": "http://localhost:3000/sse" }. OAuth mode requires creating an OAuth client and configuring a .env file; .env.gcloud-sdk-oauth can be used as the .env file.
In addition, the Cloud Run skill is based on the gcloud CLI, enabling AI agents to perform various operations on Cloud Run. Before use, you need to install and authenticate the gcloud CLI, set the project, and enable the skill in the agent AI tool. Example prompts include listing services, deploying folders, describing jobs, and so on.
Limitations and notes: the remote MCP server must use authentication, otherwise it may cause unauthorized access; the remote MCP server can only deploy code to the same Google Cloud project where it resides; the Google Cloud Platform Terms of Service and Data Processing and Security Terms do not apply to any component of the Cloud Run MCP Server software.
