MCP server

Make Artificial Analysis data part of your agentic workflow. Add the server to any MCP client, sign in, and ask.

  1. Add the server

    Paste the URL into your client's connector or MCP settings.

    Works with any MCP client that supports remote servers

  2. Sign in

    Pick an organization and allow access.

    Agent deployed on a server, where nobody can sign in? Use an API key

  3. Ask

    Your agent picks the right endpoint and answers from live data.

What your agent can read

The same read-only endpoints as the REST API, across every modality.

Connect with an API key

Signing in needs a person at a browser. For an agent deployed on a server, or a CI job, use an API key instead.

  • Create a key in API key management for the organization whose plan the agent should use.
  • Set it as AA_API_KEY where the agent runs.
  • Add the server with your client's setup below.

The server takes the key in x-api-key, or as a bearer token. Each call counts as one API request, as on the REST API; signing in gets double the usage.

APIs and platforms

Agent SDKs and frameworks

Coding agents and IDEs

Add to client.beta.messages.create

betas=["mcp-client-2025-11-20"],
mcp_servers=[{
    "type": "url",
    "url": "https://mcp.artificialanalysis.ai/mcp",
    "name": "artificial-analysis",
    "authorization_token": os.environ["AA_API_KEY"],
}],
tools=[{"type": "mcp_toolset", "mcp_server_name": "artificial-analysis"}],

Troubleshooting

Asked to sign in again and again

Remove the server and add it again.

401 invalid_token when signing in

Remove any Authorization header from the server's config.

401 with an API key

Copy the key again; keys start with aa_. If it was deleted or regenerated, create a new one.

Sign-in fails

Contact the team with the client's name.