SQL MCP Server

This project provides a Python-based Model Context Protocol (MCP) server that allows a Large Language Model (LLM) to securely interact with SQL Server or PostgreSQL databases in a read-only capacity.

Features

• Connects to PostgreSQL or SQL Server databases.
• Provides MCP tools to:
  • get_schema: Fetch database schema (all tables or a specific table).
  • query_data: Execute read-only SELECT queries.
• Enforces read-only access through query validation.

Setup

1. Environment: This project uses uv for environment and dependency management. Ensure you have uv installed (pip install uv).

2. Dependencies: The necessary dependencies have already been installed into the .venv directory using uv add. If you need to reinstall:

   uv sync
   

3. Configuration:

   • Copy the example environment file:

     cp .env.example .env
     

   • Edit the .env file and set the DATABASE_URL variable to your database connection string. Examples are provided in the file for PostgreSQL and SQL Server (using pyodbc).
   • Choose the appropriate SQLAlchemy driver prefix (postgresql+asyncpg, mssql+pyodbc, etc.) based on your database and installed driver.
   • Ensure you have the necessary database drivers installed (e.g., psycopg2-binary is included, but for SQL Server, you might need pyodbc and system-level ODBC drivers).
   • Configure the TRANSPORT (e.g., stdio or sse), HOST, and PORT as needed.

Usage

Development & Testing

You can run the server locally for testing using the MCP development tools:

# Activate the virtual environment (optional, uv run handles it)
# source .venv/bin/activate  # Linux/macOS
# .\\.venv\\Scripts\\activate  # Windows

# Run with uv (recommended)
uv run mcp dev src/main.py

# Or run directly if environment is activated
# mcp dev src/main.py

This will start the MCP Inspector, allowing you to interact with the get_schema and query_data tools.

Running Standalone

You can also run the server directly using the configured transport (stdio or sse):

# Ensure .env is configured (especially TRANSPORT)
uv run python src/main.py

Integration with MCP Clients (e.g., Claude Desktop)

Refer to the mcp install command and the MCP client's documentation for integrating the server. You'll typically provide the command to run the server (using uv run python src/main.py or similar) and necessary environment variables.

Example Stdio Configuration (Conceptual):

{
  "mcpServers": {
    "sql-explorer": {
      "command": "uv",
      "args": ["run", "python", "c:/path/to/sql-mcp/src/main.py"],
      "envFiles": ["c:/path/to/sql-mcp/.env"],
      "env": {
          "TRANSPORT": "stdio"
          // DATABASE_URL will be picked from .env file
      }
    }
  }
}

Example SSE Configuration (Conceptual):

Ensure TRANSPORT=sse, HOST, and PORT are set in .env.

{
  "mcpServers": {
    "sql-explorer": {
      "transport": "sse",
      "url": "http://localhost:8051/sse" // Or configured HOST/PORT
    }
  }
}