Summary
Schema first, then the query - the loop that usually costs two failed attempts.
It lists tables and describes them before anything runs, so column names get copied rather than guessed. Restricting it to named datasets happens at launch, which means the reach of the query tool is decided before the model ever sees it.
What it is
Three tools over a BigQuery project: the tables it can see, the schema of one of them, and a query in BigQuery dialect.
What you get
- Every table in the project, or only in the datasets you name
- The schema of a specific table
- A SQL query executed in BigQuery dialect, with rows returned inline
Requirements
A GCP project and its location - both required, given either as launch arguments or as environment variables. An optional dataset setting narrows the server to named datasets, and several can be listed.
Setup effort
One command — npx -y @smithery/cli install mcp-server-bigquery --client claude