Pular para o conteúdo
Kitbay
Servidor MCP

bigquery-mcp

Roda localmenteMITAtualizado em 26 de mai. de 202610Estrelas no GitHubExige chave de API

A local MCP server by Paul van Oorschot for exploring large BigQuery projects without filling the model's context.

Resumo

A local MCP server by Paul van Oorschot for exploring large BigQuery projects without filling the model's context. By default it returns only dataset and table names, and fetches schemas, column descriptions, fill rates and a few sample rows when asked. Its query tool accepts only SELECT and WITH statements and applies a per-query billing cap of roughly half a dollar unless changed. An optional vector search tool uses an embedding model in BigQuery to find similar rows. Access can be limited to named datasets. It is installed from PyPI with uvx and signs in with application default credentials or a service-account key.

O que faz

  • List datasets and tables with optional detail
  • Show table schema, column fill rates and sample rows
  • Run read-only SELECT queries with a billing cap
  • Run semantic similarity search on embedding tables