ContextQMD
Back to Genai Toolbox

MCP Toolbox for Databases

npm/server/README.md

1.2.02.4 KB
Original Source

MCP Toolbox for Databases

A CLI tool for running a toolbox server.

Installation

You can install the toolbox globally:

bash
npm install -g @toolbox-sdk/server

Or run it directly using npx:

bash
npx @toolbox-sdk/server

Configuration

Toolbox supports two ways to define tools: Prebuilt Sources for instant database access and Custom YAML for tailored logic.

1. Prebuilt Tools

Use the --prebuilt flag to skip configuration files and instantly expose standard database operations.

  • Usage: npx @toolbox-sdk/server --prebuilt <source> --stdio
  • Supported: See all supported datasource.
  • Setup: Requires relevant environment variables (e.g., BIGQUERY_PROJECT or POSTGRES_HOST).

2. Custom Tools (tools.yaml)

The toolbox requires a tools.yaml file in the current working directory to define sources, tools, and prompts.

  • Usage: The server loads tools.yaml automatically, or use --config <path>.
  • Function: Defines sources (connection details) and tools (specific SQL/logic with descriptions for the LLM).

Example tools.yaml

yaml
sources:
  my-pg-source:
    kind: postgres
    host: 127.0.0.1
    port: 5432
    database: toolbox_db
    user: postgres
    password: password
tools:
  search-hotels-by-name:
    kind: postgres-sql
    source: my-pg-source
    description: Search for hotels based on name.
    parameters:
      - name: name
        type: string
        description: The name of the hotel.
    statement: SELECT * FROM hotels WHERE name ILIKE '%' || $1 || '%';
prompts:
  code-review:
    description: "Asks the LLM to analyze code quality and suggest improvements."
    messages:
      - role: "user"
        content: "Please review the following code for quality, correctness, and potential improvements: \n\n{{.code}}"
    arguments:
      - name: "code"
        description: "The code to review"
        required: true

To learn more on how to configure your toolbox, visit the official docsite.

Platform Support

The toolbox automatically handles platform-specific binaries. Supported platforms include:

  • macOS (arm64, x64)
  • Linux (x64)
  • Windows (x64)

Resources

For more information, visit the