Skip to content

EggAI MCP Integration

This example demonstrates how to integrate Model Context Protocol (MCP) with EggAI, allowing you to easily connect and use MCP-compatible tools in your EggAI agents.

Overview

This example provides a way to integrate MCP tools with EggAI's agent framework, allowing agents to communicate with MCP-compatible tools through a standardized interface. The integration supports:

  1. Running MCP tools as EggAI agents
  2. Agent-to-agent communication using EggAI channels
  3. Tool discovery and usage in LLM conversations

Components

  • main.py: Main chat application that demonstrates using MCP tools with LiteLLM
  • mcp_utils.py: Utility functions for MCP integration with EggAI
  • mcp_servers/: Directory containing example MCP server implementations
  • fetch.py: Example server for HTTP fetch operations
  • filesystem.py: Example server for filesystem operations
  • sandbox/: Directory used by the filesystem MCP server

Architecture

The example follows a layered architecture:

  1. MCP Servers: Third-party tools that implement the MCP protocol (fetch and filesystem in this example)
  2. EggAI MCP Adapters: Adapters that convert MCP server communication to EggAI channels
  3. EggAI Agent: A client agent that can discover and use MCP tools through EggAI channels

Prerequisites

  • Python 3.9+
  • Kafka (provided via Docker Compose)
  • Node.js and npm (for filesystem MCP server)
  • uvx (for fetch MCP server)

Ensure you have a valid OpenAI API key set in your environment:

export OPENAI_API_KEY="your-api-key"

Setup

  1. Clone the repository and navigate to the MCP example directory: 

    cd examples/mcp

  2. Set up the environment using the provided Makefile: 

    make setup

  3. Start the Kafka infrastructure with Docker Compose: 

    make docker-up

Running the Example

You can run the services individually or all at once:

Start individual services

  1. Start the Fetch MCP service: 

    make start-mcp-fetch

  2. Start the Filesystem MCP service: 

    make start-mcp-filesystem

  3. Start the chat application: 

    make start-mcp-chat

Start all services at once

make start-all

Interacting with the Example

Once all services are running, you can interact with the chat interface. The agent has access to tools from both MCP services and can:

  • Fetch data from the web using the Fetch MCP server
  • Read and write files using the Filesystem MCP server

Example interactions: - "Can you fetch the content from https://example.com and save it to a file?" - "What files are available in the sandbox directory?" - "Read the content of file-1.txt"

Cleanup

When you're done with the example, you can clean up:

  1. Stop the Docker services: 

    make docker-down

  2. Clean up the environment: 

    make clean

Next Steps

Ready to explore further? Check out:

  • Advanced Examples: Discover more complex use cases in the examples folder.
  • Contribution Guidelines: Get involved and help improve EggAI!
  • GitHub Issues: Submit a bug or feature request.
  • Documentation: Refer to the official docs for deeper insights.