Docs
MCP

Connect the MCP server

Run the emma MCP server so Claude Desktop, Cursor, or your own agent can read and act on your infrastructure.

The emma MCP server exposes your infrastructure to any Model Context Protocol client — Claude Desktop, Cursor, or an agent you build. It runs locally as a subprocess and speaks to emma on your behalf, so your AI tools get the same grounded view of your clouds that the copilot has.

Before you start you need an emma access token. Create one in the console under Settings → API keys. Node.js (which provides npx) must be installed.

1. Get the command

The server runs via npx — no global install needed:

npx -y @emma/mcp-server

It reads your emma token from the EMMA_TOKEN environment variable (or a --token flag). Keep the token out of shared config where you can.

2. Add it to your client

Register the server with your MCP client. Most clients take a small JSON block naming the command and its environment.

Add this to your claude_desktop_config.json:

{
  "mcpServers": {
    "emma": {
      "command": "npx",
      "args": ["-y", "@emma/mcp-server"],
      "env": { "EMMA_TOKEN": "your-emma-token" }
    }
  }
}

Restart Claude Desktop, and the emma tools appear.

3. Verify the tools

Once the client reloads, the emma tools become available. Ask your client something that uses them.

Try this (in your MCP client) — List my emma resources in eu-central

The full set of tools, their parameters, and what each one does is in the MCP tools reference.

Your token scopes everything the server can see and do — treat it like a password, and revoke it in Settings → API Keys if it's ever exposed.

What's next

How is this guide?

Last updated on 17 Jul 2026

On this page