LABSCO SUMMARY
Business questions answered with SQL it writes against your BigQuery tables.
It goes from a business question to an answer in one pass, finding the right tables, reading their structure, then writing and running the SQL.
What it is
A remote MCP server Google Cloud runs at bigquery.googleapis.com/mcp. It is switched on whenever the BigQuery API is enabled in your project. It works with your Google Cloud identity and the access roles you hold.
What you get
- Datasets and tables listed in a Google Cloud project, with their details and structure
- Read-only SQL queries run and the results returned
- Any SQL BigQuery supports, including statements that insert, update, delete or create data
- Long-running queries checked on, paged through, or cancelled
Requirements
- A Google Cloud project and an identity with three roles on it: MCP Tool User, BigQuery Job User and BigQuery Data Viewer.
- You sign in with OAuth, using your Google Cloud credentials, your own OAuth client, or a separate agent identity, which Google recommends.
- Queries it runs are ordinary BigQuery jobs billed to that project.
- By default, queries are cancelled after three minutes, and results are capped at 3,000 rows.
- execute_sql can change data, so if the agent should only read, have an admin block that tool with a deny policy.
Setup effort
Paste a URL, then authorize — add the endpoint to your client, then approve the OAuth consent screen