QueryCraftMCP

QueryCraftMCP

QueryCraftMCP is a flexible Model Context Protocol (MCP) server designed to bridge the gap between Large Language Models (LLMs) and various database systems. It allows LLMs or other MCP clients to dynamically discover database schemas, execute queries, and retrieve data through a standardized protocol. The server supports multiple database backends (currently PostgreSQL and SQLite), selectable via configuration.

Features

• Multi-Database Backend Support:
  • PostgreSQL: Asynchronous interaction using asyncpg, including connection pooling via lifespan management.
  • SQLite: Synchronous interaction using the built-in sqlite3 library.
  • Easily configurable active backend using the ACTIVE_DB_BACKEND environment variable.
• Dynamic Tool Loading: Tools and database connection lifespans are dynamically loaded based on the configured backend.
• Comprehensive Database Interaction Tools:
  • Schema Discovery: Tools to list available databases (PostgreSQL), database objects (tables/views) (PostgreSQL), and object columns (PostgreSQL). For SQLite, a tool to retrieve the full table DDL schema is provided.
  • Data Querying:
    • Structured search capabilities (e.g., search_data for PostgreSQL).
    • Raw SQL query execution (e.g., execute_raw_sql for PostgreSQL, execute_query for SQLite) with security considerations.
• Lifespan Management: Robust management of database connections throughout the application lifecycle.
• Transport Protocol: Utilizes Server-Sent Events (SSE) for MCP communication.
• Configuration: Primarily through .env file and environment variables.
• Docker Support: Includes a Dockerfile (suggested in discussion, not uploaded) for easy containerization and deployment.

Project Structure

The project follows a src/ layout, with database-specific implementations organized under src/db_backends/:

Prerequisites

• Python 3.9+
• Docker (if running via Docker)
• Access to a PostgreSQL server (if using the PostgreSQL backend)
• A writable directory for the SQLite database file (if using the SQLite backend)

Setup

1. Clone the Repository:
   git clone <your-repository-url> cd QueryCraftMCP
2. Create a Virtual Environment (recommended):
   python -m venv venv source venv/bin/activate # On Linux/macOS # venv\Scripts\activate # On Windows
3. Install Dependencies:
   pip install -r requirements.txt
   The dependencies include mcp[cli], asyncpg, and python-dotenv.
4. Configure Environment Variables: Create a .env file in the project root. Populate it with the necessary configurations:
   # .env # --- General Configuration --- # Choose the active database backend: "postgres" or "sqlite" ACTIVE_DB_BACKEND="postgres" # MCP Server Host and Port (used by main.py) MCP_HOST="0.0.0.0" MCP_PORT="8888" # The port your MCP server will listen on with SSE/HTTP # --- PostgreSQL Backend Configuration --- # Required if ACTIVE_DB_BACKEND is "postgres" POSTGRES_DATABASE_URL="postgresql://your_user:your_password@your_pg_host:5432/your_database" # --- SQLite Backend Configuration --- # Required if ACTIVE_DB_BACKEND is "sqlite" # This path is relative to where the application runs. # If running in Docker, this path will be inside the container. SQLITE_DATABASE_PATH="querycraft_data.db"
   • Replace placeholder values (like your_user, your_password, etc.) with your actual credentials and paths.
   • The MCP_HOST and MCP_PORT are used by FastMCP when it's instantiated in main.py.

Running the Application

1. Locally (Directly with Python)

Ensure your .env file is configured correctly.

The server will start using the configured ACTIVE_DB_BACKEND and will listen on the host and port specified by MCP_HOST and MCP_PORT (defaulting to 0.0.0.0:8888) using SSE transport.

2. Using Docker First, build the Docker image (assuming you have a Dockerfile like the one suggested in prior discussions):

Then, run the Docker container. You'll need to pass the environment variables.

Example for PostgreSQL Backend:

• Replace docker_user, docker_pass, docker_db with your actual PostgreSQL credentials.
• host.docker.internal can be used to connect to a PostgreSQL server running on your host machine from within the Docker container (on Docker Desktop for Mac/Windows).
• -p 8888:8888 maps the host's port to the container's port.