Can AI agents automate MotherDuck?

MotherDuck is automatable today through its remote MCP server. This record is MotherDuck cloud analytics. DuckDB, ClickHouse, Snowflake, and BigQuery are separate products. Programmatic query access uses a MotherDuck access token. The token is created in the web settings as Read/Write or Read Scaling and passed as motherduck_token or in a connection string such as md:my_db?motherduck_token=<token>. A separate REST admin API at https://api.motherduck.com is marked preview. It manages service accounts, tokens, and Duckling configuration for an Admin Read/Write token sent as Authorization: Bearer. It does not list every service account in an organization. MotherDuck hosts an MCP server at https://api.motherduck.com/mcp. Clients normally use OAuth. Backend clients can send Authorization: Bearer with a MotherDuck token. Documented tool groups cover SQL, schema and shares, Dives, Flights, and Guides. Query results pass through the MCP server, which may sit outside the organization's region even though MotherDuck says the data is processed in that region. Read-only restriction is documented. The MotherDuck organization also publishes a local MCP server for DuckDB files and MotherDuck. The motherduck CLI is in preview. It supports OAuth device login, MOTHERDUCK_TOKEN, and table, json, or csv output. A dedicated MotherDuck SDK catalog was not established. Connector catalogs were not opened. https://motherduck.com/docs/llms.txt is a documentation index. Computer-use viability is unassessed. The remote MCP scores 8 and is the best path. The query token plus preview admin API scores 7. The preview CLI scores 6.

Best path today: mcp · Overall automatability: 8/10