MCP AI 助手

MCP AI assistant tutorial project

A beginner-friendly project to learn how MCP (Model Context Protocol) works with LLMs using Groq + LangChain + MCP servers.

🚀 What This Project Does

This project creates a simple AI assistant that can:

• Chat using a Groq LLM
• Use MCP tools (browser, search, etc.)
• Remember conversation history
• Run in your terminal

It demonstrates how MCP Host → Client → Server → Tool works in practice.

🧠 MCP Architecture

User → MCP Agent → LLM (Groq)
                     ↓
               MCP Client
                     ↓
               MCP Servers
                     ↓
                  Tools

• Host → Your Python app
• Client → MCPClient
• Server → Playwright / Search MCP servers
• Tools → Browser automation, search, etc.

📦 Requirements

• Python 3.10 or 3.11
• Node.js (for MCP servers)
• Groq API key
• uv or pip package manager

⚙️ Installation

1. Clone the project

git clone <your-repo-url>
cd ai-assistant

2. Create virtual environment (uv recommended)

conda create --name mcp python=3.12
conda activate mcp

3. Install dependencies

 pip install -r requirements.txt

MCP servers are started automatically from browser_mcp.json.

🔑 Setup Environment Variables

Create a .env file:

GROQ_API_KEY=your_api_key_here

▶️ Run the Project

python main.py

You will see an interactive chat like:

You: search web for python tutorial
Assistant: ...

Commands:

• exit → quit chat
• clear → clear memory

🧩 MCP Config File

Example browser_mcp.json:

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["@playwright/mcp@latest"]
    }
  }
}

You can add more MCP servers later.

⚠️ Common Issues

1. ModuleNotFoundError

Fix import paths:

from langchain_groq import ChatGroq

2. Search or Airbnb MCP fails

Some MCP servers use scraping and may be blocked. Use official APIs instead.

3. Token limit errors

Reduce:

• number of MCP tools
• conversation memory
• max_steps