Connect

Connect an AI agent

This is about connecting an AI coding agent to the Aaly platform itself — so it can define entities and fields through MCP tools. For your app's own end-user login, see Auth.

One-click OAuth (recommended)

Add this URL as a custom MCP connector in Claude, Cursor, or Codex:

https://mcp.aaly.io

That's the whole configuration. Approving the consent screen that opens in your browser completes the rest:

  1. Your agent requests the server and gets a 401 with a WWW-Authenticate: Bearer resource_metadata=... challenge (RFC 9728).
  2. It fetches /.well-known/oauth-authorization-server (RFC 8414) to discover the authorization and token endpoints.
  3. It registers itself dynamically via the registration_endpoint (RFC 7591) — nothing to pre-provision in a dashboard.
  4. It runs an OAuth 2.1 authorization-code flow with PKCE (S256), opening the consent screen in your browser.
  5. It holds a scoped, revocable token. No key is ever shown to you or pasted by you.

Verified against the live server:

$ curl -i https://mcp.aaly.io
HTTP/1.1 401 Unauthorized
www-authenticate: Bearer resource_metadata="https://mcp.aaly.io/.well-known/oauth-protected-resource"

Tested end to end against Claude and Codex.

Revoking access

Revoke a connected agent's token from the dashboard at any time — it's scoped per agent, so revoking one doesn't affect any other connector or any API key you've issued separately.

API key (advanced)

For CI, scripts, server-side integrations, and any client that can't open a browser. Generate a key in the dashboard, or via POST /auth/keys on your project's REST API (see Auth for the exact request), then configure your MCP client with it directly:

{
  "mcpServers": {
    "aaly": {
      "type": "http",
      "url": "https://mcp.aaly.io",
      "headers": { "Authorization": "Bearer aaly_<id>_<secret>" }
    }
  }
}

Same server, same tools, same tenant scoping as the OAuth path — you just manage the credential yourself. Revoke it with DELETE /auth/keys/{id} or from the dashboard.

What your agent can do once connected

Every MCP tool call operates on your project's definition — entities, fields, and schema inspection — never on your app's end-user data directly. Full list: MCP tools reference.