Connect an AI client to the Yatmo MCP server

The Yatmo MCP server speaks the standard Streamable HTTP MCP transport. Any AI client that supports a remote MCP server can connect with a tiny JSON snippet and one HTTP header.

Generic remote-MCP config

Most clients accept this exact shape (sometimes nested under a slightly different parent key — the URL + headers parts are universal):

{
  "mcpServers": {
    "yatmo": {
      "url": "https://mcp.yatmo.com/mcp/v1",
      "headers": { "LicenceKey": "YOUR_KEY" }
    }
  }
}
Where the key lives
Put your LicenceKey in the client’s secret store, not in source control. Most clients accept environment-variable substitution; check your client’s docs.

Claude.ai / Claude Desktop

In Claude.ai, add the Yatmo MCP server as a custom Connector with:

The four tools (yatmo_get_location_summary, yatmo_get_nearby_pois, yatmo_get_nearest_by_category, yatmo_get_accessibility_profile) appear in the connector list once it’s enabled.

Cursor and other IDE / agent clients

Cursor and similar tools accept the generic shape in their MCP configuration file. The exact filename varies; the snippet does not:

{
  "mcpServers": {
    "yatmo": {
      "url": "https://mcp.yatmo.com/mcp/v1",
      "headers": { "LicenceKey": "YOUR_KEY" }
    }
  }
}

Custom AI agents

If you are building an agent directly against the Anthropic, OpenAI or another LLM SDK:

See the quick start for a raw HTTP example.

Browser-side clients (don’t)

Do not connect from a public web page. The LicenceKey grants quota-bearing access to your Yatmo account; exposing it in browser-side JavaScript leaks it to every visitor.

If you need Yatmo data on a public website, use the Yatmo plugins (which use a frontend-only key with stricter limits) or proxy the MCP server through your own backend.

Related: Authentication & error codes, Quick start, FAQ.