Connect Claude, Cursor, or any MCP client and let your agent search, inspect and manage your tenant on your behalf.
There's no hosted endpoint to connect to: the server runs locally over stdio. The MCP guide covers setup, capabilities and how the server is generated from the same OpenAPI contract as the REST API.
claude mcp add aacsearch --env API_KEY_API_KEY_HEADER=your-api-key -- node /path/to/aacsearch/sdks/mcp/build/index.jsPick your agent to register the built server as a local command, or copy the config by hand. Every client is covered in the full MCP server docs.
Run a search or multi-search against any collection and inspect the raw response.
List, inspect, create, and update collection schemas.
Look up, index, update, and delete documents; import or export in bulk.
Read and edit merchandising rules, synonyms, and curation for a collection.
List recommendation models and fetch recommendations for a given anchor.
Read search analytics: top queries, zero-result rate, click-through rate.
List experiments and their reports; start, stop, or promote a variant.
Run a golden query set and read back nDCG, MRR, MAP and precision/recall.
See the available tools reference for every tool name and parameter.
Generate an API key from the console's Settings → API Keys, and pass it to the local server process via the API_KEY_API_KEY_HEADER environment variable in the MCP client config. Scope it to the collections and operations your agent actually needs — see the authentication docs.