MCP Toolbox for Databases
googleapis/mcp-toolboxReviewed by AI
Checked by RepoPicks Radar on
What is MCP Toolbox for Databases?
An open-source MCP server, published by Google, that sits between an AI assistant (Gemini CLI, Claude Code, Codex, Cursor and others) and databases such as PostgreSQL, MySQL, SQL Server, Oracle, MongoDB, Redis, BigQuery or Spanner. Ready-made tools let the assistant list tables and run queries, and a tools.yaml file lets developers define their own, narrower tools. It is for developers and data teams: you need a terminal, a database you already have access to and a configuration file. Apache-2.0 licensed and actively developed.
How easy is it to install MCP Toolbox for Databases?
D · terminal, several tools or a specific graphics card · Windows
- Windows D
- macOS D
- Linux D
The same terminal work on Windows, Mac and Linux: the featured quick start runs `npx @toolbox-sdk/server` (so Node.js), or you download the toolbox binary with curl, and in both cases you still write a tools.yaml with database connection details and paste an MCP block into your assistant's settings. Homebrew helps on Mac and Linux, but there is no installer and nothing works until you already have a database to point at.
Coming soon: install in 1 click with the RepoPicks app.
MCP Toolbox for Databases on GitHub
- Stars
- 16,835on Nov 10, 2026
- Downloads
- —all releases
- Growth
- +35stars in 6 days
Stars over time
Facts
- Runs on
- Windows, macOS, Linux
- You need to install
- Node.js
- To build from source
- Docker, Go
- Disk space
- not determined
- Accounts or keys
- Database credentials (user and password or Google Cloud IAM), Google Gemini (GEMINI_API_KEY / GOOGLE_API_KEY, only for the example apps in the README)
- Cost
- Free · The server is free and open source. The databases or Google Cloud services you point it at may cost money, and the AI assistant you use it with may be paid.
- License
- Apache-2.0 (Permissive)
- Last change on GitHub
- Nov 10, 2026
MCP Toolbox for Databases risk signals (3)
- Attentionai can run queries on your database
README: "Prebuilt generic tools for instant data exploration (e.g., `list_tables`, `execute_sql`) directly from your IDE or CLI." With the prebuilt toolsets the AI assistant can send SQL to the database you connect, so the account you give it decides what it can read or change.
AI review
- Infocredentials in plain config file
The README's source example puts the database password directly in the config file: "user: toolbox_user" / "password: my-password" in tools.yaml.
AI review
- Infomany open issues
431 open issues on GitHub, spread over dozens of supported databases, so support for any one database may be uneven.
AI review
Evidence is shown in its original language.
What Radar could not check about MCP Toolbox for Databases
- Whether each of the listed databases really works, since each has its own driver
- Whether the prebuilt toolsets can be limited to read-only in practice
- Disk and memory use of the server once running
- What the AI assistant you connect does with the data it reads from your tables