MCP Server with SQL Alchemy and Gemini LLM

This project implements a Model Context Protocol (MCP) server that provides a bridge between a PostgreSQL database and an AI agent powered by Google's Gemini LLM. The server allows natural language interactions with the database through MCP endpoints.

Features

• SQL Alchemy integration for database operations
• MCP Server implementation with standardized endpoints
• Gemini LLM-powered agent for natural language database interactions
• PostgreSQL database support
• Structured query processing and response handling

Project Structure

mcp-server-client/
├── main.py              # Main application entry point
├── pyproject.toml       # Project dependencies and configuration
├── client/             
│   ├── __init__.py     
│   └── db_client.py    # Database client implementation
├── server/             
│   ├── __init__.py    
│   ├── postgres_db.py  # PostgreSQL database connection and operations
│   └── weather.py      # Weather-related functionality

Installation

1. install uv:

pip install uv

2. activate virtual environment'

# for linux
source .venv/bin/activate

# for windows
.venv\Scripts\activate

3. Install dependencies:

uv add mcp[cli] mcp-use psycopg2 psycopg2-binary sqlalchemy langchain-google-genai

4. Set up environment variables:

# create .env file
touch .env

# add this into .env file
# PostgreSQL connection
DB_URL= "postgresql://postgres:admin@localhost:5432/mydb"
GEMINI_API_KEY= "API_KEY"

5. start the agent

uv run client/db_client.py 

Example Interactions

Below are some examples of interactions with the SQL agent: