lokeshtalamala1/mcp-server-project

MCP Server Project (Local)

This project demonstrates a complete local MCP server + client architecture for querying a PostgreSQL banking database via AI agent tools using OpenAI.

📁 Project Structure

mcp-server-project/
├── mcp_server.py        # MCP server with database tools
├── mcp_client.py        # Interactive agent with OpenAI LLM
├── server.json          # MCP client config for launching the server
├── .env                 # API key for OpenAI (DO NOT COMMIT)
├── requirements.txt     # Project dependencies
└── README.md            # Setup and usage instructions

✅ Prerequisites

• Python 3.10+
• PostgreSQL running locally with the correct schema
• Access to OpenAI API key (put it in .env)

🔧 Setup Instructions

1. Clone or Copy the Folder

2. Create Virtual Environment

   python -m venv .venv
   source .venv/bin/activate   # On Windows: .venv\Scripts\activate
   

3. Install Dependencies

   pip install -r requirements.txt
   

4. Set Up .env File Create a .env file in the root:

   OPENAI_API_KEY=your-api-key-here
   

🚀 Running the Project

1️⃣ Start the MCP Client (automatically launches server)

python mcp_client.py

• This will:
  • Load server.json
  • Start mcp_server.py
  • Connect to it via MCPClient
  • Create a memory-enabled LLM agent
  • Accept user queries (e.g., "Get transactions for customer 12345")

🛠 Available Tools

Your LLM agent has access to the following MCP tools:

1. get_customer_transactions(customer_id, months_back)
2. classify_and_store_transactions(customer_id, months_back)
3. get_mcc_details(mcc_code)
4. get_tc_details(tc_code)

📦 Database Config

Database credentials are pulled from .env. Update them as needed:

DB_HOST=localhost
DB_PORT=5432
DB_NAME=mcp_db
DB_USER=postgres
DB_PASS=lokit@181903

Update the mcp_server.py accordingly if you want to use dynamic environment variables.

🔄 Transition to Remote Server (Later)

This setup is for local development only. Once tested, you can:

• Deploy mcp_server.py to AWS EC2 (or other cloud)
• Update the server.json to point to a remote server URL
• Keep mcp_client.py on your local machine to communicate with it