MCP Catalogs
← Home

mcp-bigquery-server

by ergut·★ 139·Score 48

A secure MCP server enabling LLMs to query BigQuery datasets with field-level data protection.

databaseai-llmsecurity
33
Forks
1
Open issues
1 mo ago
Last commit
2d ago
Indexed

Overview

This MCP server provides secure, read-only access to Google BigQuery datasets through a natural language interface. It allows LLMs to analyze data without writing SQL directly, while offering robust protection for sensitive information. The server operates in two modes: Simple Mode for personal projects with non-sensitive data, and Protected Mode for environments containing PHI, PII, or financial data where field-level access restrictions are critical.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

you:Natural language analysis of business data stored in BigQuery
you:Healthcare data analysis with HIPAA compliance through field-level restrictions
you:Financial reporting with automatic protection of sensitive columns
you:How does this server protect sensitive data?
you:Which LLM interfaces are supported?

When to choose this

Choose this MCP server if you're already using Google BigQuery and need AI-powered data access with sensitive field protection capabilities.

When NOT to choose this

Don't choose this if you need write access to BigQuery or are looking for a database-agnostic solution that works with multiple cloud data warehouses.

Tools this server exposes

3 tools extracted from the README
  • query_bigquery

    Execute SQL queries against BigQuery datasets with configurable safety limits

  • get_table_schema

    Retrieve schema information for BigQuery tables and views

  • list_tables

    List available tables and views in BigQuery datasets

Comparable tools

bigquery-shell-mcpduckdb-mcppostgres-mcp

Installation

Installation

  1. **Install prerequisites**: Node.js 14+, Google Cloud project with BigQuery enabled, and Claude Desktop
  1. **Authenticate with Google Cloud**:

``bash gcloud auth application-default login ``

  1. **Configure Claude Desktop** (claude_desktop_config.json):

``json { "mcpServers": { "bigquery": { "command": "npx", "args": [ "-y", "@ergut/mcp-bigquery-server", "--project-id", "your-project-id" ] } } } ``

For Protected Mode with sensitive data, add --config-file /path/to/config.json to the args.

FAQ

How does this server protect sensitive data?
The server implements field-level restrictions that prevent AI agents from accessing sensitive columns. It can auto-discover sensitive fields using patterns or allow manual configuration of prevented fields. When a query tries to access restricted data, the agent receives guidance to reformulate using aggregates or EXCEPT clauses.
Which LLM interfaces are supported?
Currently, only Claude Desktop is officially supported as it has MCP capabilities. The server is designed with the Model Context Protocol but may work with other MCP-compatible interfaces in the future.

Compare mcp-bigquery-server with

mcp-bigquery-server vs ultimate_mcp_servermcp-bigquery-server vs mcp-server-chartmcp-bigquery-server vs everythingmcp-bigquery-server vs memorymcp-bigquery-server vs fetch
GitHub →

Last updated · Auto-generated from public README + GitHub signals.