MCP Knowledgebase Llm

🚀 MCP Knowledge Base

A simple MCP client-server system

🚀 Quick Start

✨ Features

This project is a simple MCP client - server system that allows for direct tool calls and interactions with an LLM using the OpenAI API.

📦 Installation

• Python 3.9 or higher
• Poetry for dependency management
• An OpenAI API key

📦 Installation Steps

1. Install dependencies using Poetry:

poetry install

2. Create a .env file in the project root directory or its parent directory, containing your OpenAI API key:

OPENAI_API_KEY=your_api_key_here

📚 Documentation

Project Structure

• server.py: Implements the MCP server with tool - related functionality.
• client - sse.py: Implements the MCP client with LLM - related functionality.
• data/kb.json: Knowledge base data containing MCP - related Q&A.
• pyproject.toml: Poetry configuration file.

Running the Application

1. Start the server:

poetry run python server.py

2. Run the client in another terminal:

poetry run python client - sse.py

Using the Client

The client has two modes:

1. Direct tool call:

   • Uncomment the line asyncio.run(test_direct_tool_calls()) in client - sse.py.
   • This will directly call the tool without using the LLM.

2. Interaction with LLM (default):

   • Use OpenAI to interpret queries and call appropriate tools.
   • Ask questions such as "What is MCP?" or "What's the difference between stdio and SSE transport?"

Customization

• Add new tools in server.py by creating additional functions with the @mcp.tool() decorator.
• Modify the knowledge base by updating data/kb.json.
• Change the OpenAI model by modifying the model parameter in the MCPClient class.