genai-toolbox

The MCP Toolbox for Databases (GenAI Toolbox) provides a server that exposes a suite of database-oriented GenAI tools and integrations. It acts as a bridge between your agents or copilots and your data sources, handling concerns like connection management, authentication, and tool orchestration so your AI workflows can focus on generating useful data-driven results. Once running, you can load a tools configuration (tools.yaml) that defines the available database tools, prompts, and toolsets, enabling agents to query schemas, generate SQL, manage schemas, and run data tasks in a structured, observable way.

To use it, start the toolbox server (via npx as shown in the setup) and supply your tools.yaml configuration. The server will expose the defined tools to your MCP-based agents or Gemini/IDEs integrations, allowing natural language prompts to be translated into database actions, code generation, and data retrieval. You can integrate Toolbox with Gemini CLI extensions or other MCP clients to orchestrate complex tool sequences, manage authentication, and monitor tool performance through built-in observability hooks.

Prerequisites:

• Node.js and npm installed on your system (recommended latest LTS).
• Git installed (optional for cloning booklets or templates).

Install and run:

1. Install Node.js and npm from https://nodejs.org if not already installed.
2. Run the MCP Toolbox server using NPX (no persistent install required):

npx -y @toolbox-sdk/server --tools-file tools.yaml

3. If you prefer a persistent installation, you can install the package locally or globally and run the server from a script. For example:

# Install locally
npm install @toolbox-sdk/server

# Run (adjust path to your local node_modules if needed)
npx @toolbox-sdk/server --tools-file tools.yaml

4. Prepare your tools.yaml configuration that defines Sources, Tools, Toolsets, and Prompts, as described in the Toolbox documentation. Then point the server to that file with --tools-file or via the corresponding environment/config mechanism supported by your deployment method.

Notes and tips:

• The toolbox is designed to be used in non-production experiments via npx; for production deployments, set up a proper installation and run as a service.
• If you’re deploying in environments without interactive prompts, supply a tools.yaml via the --tools-file flag or the TOOLS_FILE environment variable.
• Ensure database connection credentials and network access are configured securely (use environment-based secrets management in production).
• Check the Toolbox docs for details on configuring Sources (databases), Tools (specific actions like query generation, schema introspection), Toolsets (groupings of tools), and Prompts (instructions to control AI behavior).
• Monitor observability and metrics integration (OpenTelemetry) as part of your deployment to keep track of tool performance and reliability.