DuckDuckGo Search with MCP Agent

This project demonstrates how to use DuckDuckGo MCP Server with a LangChain Groq LLM agent to perform intelligent search tasks via MCP (Micro Component Protocol).

Features

• MCP Server Integration (DuckDuckGo search)
• Groq LLM (deepseek-r1-distill-llama-70b) for reasoning
• Async Python execution
• Simple and modular

Installation

1. Clone the repository:

git clone https://github.com/alihassanml/Duckduckgo-with-MCP.git
cd Duckduckgo-with-MCP

2. Install dependencies:

pip install -r requirements.txt

(Include libraries like langchain_groq, python-dotenv, etc. in your requirements.txt.)

3. Set up your .env file:

GROQ_API_KEY=your_groq_api_key_here

4. Install the MCP Server:

uvx -y duckduckgo-mcp-server

(Make sure uvx is installed. If not, install it.)

Usage

Run the main script:

python main.py

This will:

• Start the MCP client
• Connect to the duckduckgo-mcp-server
• Use the Groq LLM to perform a smart search
• Print the result

Example Code

import asyncio
import os
from dotenv import load_dotenv
from langchain_groq import ChatGroq  
from mcp_use import MCPAgent, MCPClient

async def main():
    load_dotenv()
    config = {
        "mcpServers": {
            "ddg-search": {
                "command": "uvx",
                "args": ["-y", "duckduckgo-mcp-server"]
            }
        }
    }
    client = MCPClient.from_dict(config)
    llm = ChatGroq(model="deepseek-r1-distill-llama-70b")
    agent = MCPAgent(llm=llm, client=client, max_steps=30)
    result = await agent.run("Find the best restaurant in San Francisco")
    print(f"\nResult: {result}")

if __name__ == "__main__":
    asyncio.run(main())

Resources

• DuckDuckGo MCP Server
• LangChain Groq Documentation
• Micro Component Protocol (MCP)

License

This project is licensed under the MIT License.