Connect an AI client

Attlaz exposes your connections as MCP tools, so an AI client — Claude, Cursor, or any other MCP-compatible client — can use your connected systems directly.

Access is scoped to one connection. You connect a client to a single connection, and the tools it gets are the operations that connection's connector exposes. A client that needs two connections is configured with two URLs.

What you need

• A working connection — see Create your first connection.
• Its connection ID.
• An access token requested with the MCP resource as its audience. A token issued without it is rejected on /mcp/* with 403 Forbidden. See MCP § Authentication.

Claude Code

Add to your MCP configuration (.mcp.json or claude_desktop_config.json):

{
  "mcpServers": {
    "attlaz": {
      "type": "streamableHttp",
      "url": "https://api.attlaz.com/mcp/{connectionId}",
      "headers": {
        "Authorization": "Bearer {TOKEN}"
      }
    }
  }
}

Replace {connectionId} with the ID of your connection and {TOKEN} with your Attlaz access token.

Other MCP clients

Any MCP client that supports Streamable HTTP transport can connect. Configure it with:

• URL: https://api.attlaz.com/mcp/{connectionId}
• Method: POST
• Header: Authorization: Bearer {TOKEN}
• Content-Type: application/json

Spec-compliant clients can discover the authorization server themselves: an unauthenticated request to any /mcp/* endpoint returns 401 with a WWW-Authenticate header pointing at the protected-resource metadata.

Check it works

Ask the client to list its tools, or call tools/list directly. You should see the operations of the connector behind your connection — see MCP tools.

Next steps

• MCP — the full scoping model, OAuth details and protocol reference.