MotherDuck connector

MotherDuck is cloud DuckDB. It uses the same duckdb driver and the same SQL — you just point at an md: database and authenticate with a service token. So this connector is a thin subclass of the DuckDB connector: same engine, same resilience, plus your cloud-hosted (and shared) databases.

# sources.yaml
cloud:
  type: motherduck
  database: my_db                # optional — omit to attach all your databases
  token: ${MOTHERDUCK_TOKEN}     # optional — ${ENV_VAR} expansion supported

Key	Purpose
`database`	MotherDuck database name (becomes the `md:<database>` target). Omit for the bare `md:` connection, which attaches all your databases.
`token`	MotherDuck service token. `${ENV_VAR}` is expanded. Omit to use the `motherduck_token` environment variable DuckDB reads on its own.
`duckdb_config`	Optional extra settings passed to `duckdb.connect` (e.g. `custom_user_agent`).

Once connected, query it like any DuckDB source — database-qualify a table if you've attached more than one:

SELECT region, sum(amount) AS revenue
FROM my_db.sales
GROUP BY region
ORDER BY revenue DESC

Resilience: inherited from the DuckDB connector — if a query invalidates the connection, query() rebuilds it and retries once, so one bad query can't break every later query on the long-lived connection.

Note

A ${param} always substitutes a quoted string literal (injection-safe — see Queries), exactly as with the local DuckDB connector.

Extra: none — in the core install. The duckdb core dependency ships the MotherDuck extension, which auto-loads on the first md: connect.