MCP/MCP Toolbox for Databases

MCP Toolbox for Databases

Google open-source database MCP server and custom tool framework supporting BigQuery, Cloud SQL, Postgres, MySQL, Spanner, and 20+ DBs.

rag-data•Data, Analytics & BI

GitHub16.7k

License:Apache-2.0

OFFICIAL

@Google

About this Server

MCP Toolbox for Databases (`mcp-toolbox` / `googleapis/mcp-toolbox`, formerly Gen AI Toolbox) is Google’s open-source database MCP server and custom tools framework. It gives AI coding agents prebuilt and custom structured tools for schema exploration, NL2SQL execution, semantic search, and parameterized query execution across BigQuery, Cloud SQL, AlloyDB, Spanner, Firestore, PostgreSQL, MySQL, ClickHouse, Snowflake, and MongoDB with IAM auth and OpenTelemetry observability.

Real-time Interaction

Direct low-latency RPC protocol communication with structured JSON schemas.

Scoped Sandboxing

Restricted capability bounds to prevent accidental modifications or leakage.

Security & Best Practices

Restricted Query Boundaries: Prebuilt and custom tools execute parameterized statements, preventing raw injection hazards.

Native Cloud IAM Authentication: Supports Google Cloud IAM, workload identity federation, and connection pooling.

Audited Run-Time Observability: Full OpenTelemetry metrics and traces emitted for all database tool invocations.

When to Use (and When Not To)

Best For

•Querying enterprise databases like BigQuery, Cloud SQL, AlloyDB, Postgres, and Snowflake from AI assistants
•Defining hardened, deterministic database tools via tools.yaml manifests for production agent workflows
•Natural language to SQL exploration with automated schema inspection and table discovery

Not For / Limitations

•Observability metrics, PromQL, and Loki log inspection (use Grafana MCP Server instead)
•Public mobile app store user feedback and sentiment benchmarking (use App Store Review Intelligence MCP instead)
•SEO rank tracking and SERP performance analytics (use AltoRank MCP instead)

Install & Setup

One-line terminal command

terminal command

$claude mcp add toolbox-postgres -- npx -y @toolbox-sdk/server --prebuilt=postgres --stdio

Exposed Tools & Endpoints

6 tools

Functions your AI agent can invoke via this Model Context Protocol server

list_tables• callable tool

Lists all tables, views, and schemas available in the configured database connection.

describe_table• callable tool

Retrieves the column definitions, primary keys, data types, and index constraints for a specific table.

execute_sql• callable tool

Executes parameterized SQL queries against the connected database engine with safety limits.

search_database• callable tool

Performs semantic vector search or full-text query matching across configured database collections.

list_datasets• callable tool

Discovers analytical datasets and partitioned tables in data warehouses such as BigQuery and Snowflake.

load_toolset• callable tool

Dynamically loads a specific named toolset group from the toolbox configuration manifest.