@peaka/mcp-server-peaka package runs the server over stdio. Your MCP client launches it on demand with npx, and authentication is a single environment variable.
This is the right choice when you want:
- A locally-running server, no remote dependency.
- API-key auth (for CI, headless scripts, or pinning to a specific Peaka identity).
- Faster cold-starts than a remote round-trip.
Configure your client
Add the server to your MCP client config:<YOUR_API_KEY> with a Peaka API key. Restart the client after editing the config.
Supported clients shows where each client’s config file lives.
Get an API key
You can use either a Project or Partner API key:- Project API key — bound to a single Peaka project. Simplest. Good for everyday use against one project. See How to generate API keys.
- Partner API key — enumerates projects across organizations and workspaces. Use this if you want the assistant to discover and switch between projects. See How to manage Partner API keys.
Environment variables
Verify it’s working
After your client restarts:- Confirm the
peakaserver shows up in your client’s MCP server list. - Ask your assistant: “List my Peaka projects.”
- The first response should call
peaka_list_projectsand return the projects your key can see.
Updating
Because the config uses@latest, every cold start picks up the newest server release. To pin a version: