MCP-Use

MCP-Use is an open-source client library designed to seamlessly connect any Large Language Model (LLM) to various MCP (Multi-Channel Protocol) tools, enabling the creation of custom agents with access to diverse functionalities such as web browsing and file operations. This library empowers developers to build applications without relying on closed-source or proprietary application clients.

Key Features

• Versatile Integration: Easily connect any LLM to tools like web browsers, file systems, and more, facilitating advanced computing capabilities.
• Customization: Build agents tailored to specific tasks using a straightforward and flexible API.
• Multi-Server Support: Configure the library to work with multiple MCP servers concurrently, enhancing operational efficiency.

Getting Started

MCP-Use can be installed via pip, allowing for quick setup and integration into existing projects. Developers can also clone the repository to install from source.

Example Usage

To quickly demonstrate the capabilities of MCP-Use, initiate an agent that performs a web search:

import asyncio
import os
from dotenv import load_dotenv
from langchain_openai import ChatOpenAI
from mcp_use import MCPAgent, MCPClient

async def main():
    load_dotenv()  # Load environment variables
    client = MCPClient.from_config_file("browser_mcp.json")  # Initialize MCP client
    llm = ChatOpenAI(model="gpt-4o")  # Define the language model
    agent = MCPAgent(llm=llm, client=client, max_steps=30)  # Create the agent
    result = await agent.run("Find the best restaurant in San Francisco USING GOOGLE SEARCH")  # Execute query
    print(f"\nResult: {result}")

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

Example Configuration File

The following is an example of a configuration file used for initiating an MCP Client:

{
  "mcpServers": {
    "playwright": {
      "command": "npx",
      "args": ["@playwright/mcp@latest"],
      "env": {
        "DISPLAY": ":1"
      }
    }
  }
}

Additional Use Cases

MCP-Use also supports a variety of other use cases like:

• Web Browsing with Playwright: Execute web searches and navigate websites programmatically.
• Airbnb Search: Automate searches for accommodation based on user-defined criteria.
• Blender 3D Creation: Interface with 3D modeling tools for creative applications.

Configuration Support

MCP-Use allows for the initialization of sessions from configuration files, making it easier to manage different server setups:

import asyncio
from mcp_use import create_session_from_config

async def main():
    session = create_session_from_config("mcp-config.json")  # Create a session from config
    await session.initialize()  # Initialize the session
    await session.disconnect()  # Disconnect when done

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

Future Enhancements

The roadmap includes plans to support multiple servers simultaneously and test various remote connectors, expanding the functionality of MCP-Use.

Contributing

Contributions are welcome. Developers are encouraged to report bugs and submit feature requests to enhance the library's capabilities.

Requirements

• Python 3.11+
• An MCP implementation like Playwright MCP
• LangChain and appropriate model libraries (like OpenAI, Anthropic)

Citation

For academic use, cite MCP-Use as follows:

@software{mcp_use2025,
  author = {Zullo, Pietro},
  title = {MCP-Use: MCP Library for Python},
  year = {2025},
  publisher = {GitHub},
  url = {https://github.com/pietrozullo/mcp-use}
}

Explore MCP-Use to enhance the capabilities of your applications with comprehensive access to tools and resources.