Where it fits
Cloud engineers who want agents to handle Compute Engine deployment and operations work.
- List the running VMs in this project
- Start a stopped instance before a deployment
- Check which disks are attached to a machine
A remote MCP server advertises tools to your client. The provider controls the tools, the data they can reach, and the permissions they need. Read the provider's docs before you use it with private data.
Skills that pair with Google Compute Engine
These skills cover adjacent tasks, matched by purpose and the work each one describes.
Use this skill when the user wants to deploy an agent, deploy an ADK agent, set up CI/CD, configure secrets, or troubleshoot a deployment, or needs guidance on Agent Runtime, Cloud Run, or GKE deployment targets. It covers deployment workflows, service accounts, rollback, and production infrastructure. Part of the Google ADK (Agent Development Kit) skills suite. Do not use it for API code patterns (use google-agents-cli-adk-code) or evaluation (use google-agents-cli-eval).
Use this skill when the user wants to set up tracing, monitor an agent, configure logging, add observability, or debug production traffic, or needs guidance on monitoring deployed agents, including ADK (Agent Development Kit) agents. It covers Cloud Trace, prompt-response logging, BigQuery Agent Analytics, third-party integrations (AgentOps, Phoenix, MLflow, and others), and troubleshooting. Part of the agents-cli skills suite. Do not use it for deployment setup (use google-agents-cli-deploy) or API code patterns (use google-agents-cli-adk-code).
Use this skill when the user wants to create an agent project, start a new ADK project, build a new agent, add CI/CD to a project, add deployment, enhance a project, or upgrade a project. Part of the agents-cli skills suite. It covers the `agents-cli scaffold create`, `scaffold enhance`, and `scaffold upgrade` commands, template options, deployment targets, and the prototype-first workflow. Do not use it for writing agent code (ADK projects: use google-agents-cli-adk-code) or deployment operations (use google-agents-cli-deploy).
No skills found
Try a different search term or clear the filters.
Install it in your client
Choose a client, run the command or use the one-click install, then verify the connection. The endpoint uses Streamable HTTP.
- 01
Run this command in your terminal
This adds Google Compute Engine as a remote Streamable HTTP server in Claude Code.
Claude Codeclaude mcp add --transport http google-compute-engine https://compute.googleapis.com/mcp --scope project - 02
Verify the connection
Run claude mcp list, or open /mcp in a Claude Code session.
Learn more about MCP in Claude Code
- 01
Run this command in your terminal
This adds Google Compute Engine as a remote Streamable HTTP server in Codex.
Codex# Add this to Codex MCP settings: [mcp_servers."google-compute-engine"] url = "https://compute.googleapis.com/mcp" - 02
Verify the connection
Run codex mcp list, or open /mcp in a Codex session.
Learn more about MCP in Codex
- 01
Install with one click
This opens Cursor with the server pre-filled. Review the connection before accepting it.
Install in Cursor - 02
Verify the connection
Open Cursor's MCP settings and enable the server if it is not active.
Learn more about MCP in Cursor
- 01
Install with one click
This opens VS Code with the server pre-filled. Review the connection before accepting it.
Install in VS Code - 02
Verify the connection
Run MCP: List Servers from the Command Palette and start the server.
Learn more about MCP in VS Code
More clients, including project and global config files, are in the sidebar.
Endpoint and configuration
The installer uses the server name google-compute-engine. If you add the endpoint by hand, keep the same name so the guide and your config match.
Check the connection
After you add the server, confirm that the client can see its tools and that one small request returns what the provider describes.
- 1Open the tool list.
Restart the client or start a new session. Confirm that
google-compute-engineappears and that the tools match the provider documentation. - 2Run one small request.
Start with: List the running VMs in this project. Check the result before you give the server more access.
- 3Check permissions.
Review the OAuth scopes or API-key permissions, then remove the server and revoke its token when you no longer need it.
We checked the provider's page, not the MCP connection itself. No MCP sign-in or tool call was attempted. The provider page was available when we checked.
Common failure modes
These are the problems to check first. If the provider changes its endpoint or authentication flow, use its documentation as the source of truth.
- The sign-in window closes without returning. Allow pop-ups and finish the provider login in the same browser.
- The server connects but its tools are missing. Remove it, add it again, and approve the scopes the provider lists.
- Your client does not support Streamable HTTP. Use a client that does, and confirm the provider's current endpoint and transport in its docs.
Before you connect
- The provider can see the requests and data your agent sends through the server.
- Review OAuth scopes and API-key permissions. Start read-only if the provider offers it.
- Use project scope for repository-specific servers; use global scope for personal tools.
- Remove the server and revoke its token when you stop using it.
Check the connection before a real task
These guides cover the decisions that a provider description cannot answer: endpoint ownership, scopes, tool access, and a read-only test.
Provider documentation
Setup requirements, available tools, and rate limits live with the provider. Use the source link as the authority for this server.
Google Compute Engine documentationhttps://claude.com/connectors/google-compute-engine