AutoGen Extension for MCP Tools

🚨 This repository is now archived 🚨

The functionality of this project has been integrated directly into the official autogen-ext package and is available from version v0.4.6.
You can now use it directly in autogen-ext without needing this separate package.

For further updates and improvements, please refer to the official autogen repository.

About

This package provides integration between Microsoft AutoGen and the Model Context Protocol (MCP), enabling AutoGen agents to seamlessly connect with various data sources and tools.

MCP is an open standard that enables secure, two-way connections between AI systems and data sources, replacing fragmented integrations with a single universal protocol. This allows AI assistants to maintain context as they move between different tools and datasets.

Installation

pip install autogen-ext-mcp

Features

• Seamless integration of MCP tools with AutoGen agents
• Connect to any MCP-compatible data source or tool
• Automatic conversion of MCP tool schemas to AutoGen-compatible formats

Quick Start

from autogen_ext_mcp.tools import mcp_server_tools, StdioServerParams
from pathlib import Path

# Get desktop path cross-platform
desktop_path = str(Path.home() / "Desktop")

# Connect to FileSystem MCP server
server_params = StdioServerParams(
    command="npx",
    args=[
        "-y",
        "@modelcontextprotocol/server-filesystem",
        desktop_path,  # Allow access to Desktop directory
    ]
)

# Get tools
tools = await mcp_server_tools(server_params)

# Use tools with AutoGen agents
# The tools can be passed to any AutoGen agent that supports tool use

Samples

Check out the samples directory for example code and detailed instructions on getting started with the AutoGen MCP Extension. The samples demonstrate common use cases and best practices for integrating MCP tools with AutoGen agents.

Resources

• AutoGen Documentation
• Model Context Protocol Documentation
• MCP Quickstart Guide
• Pre-built MCP Servers

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.