Agents MCP Usage

🚀 Model Context Protocol (MCP) Proxy Framework Demonstration

This repository showcases the usage of a simple Model Context Protocol (MCP) server with multiple frameworks:

• Google Agent Development Kit (ADK)
• LangGraph Agents
• OpenAI Agents
• Pydantic-AI

✨ Features

Model Context Protocol (MCP)

The Model Context Protocol enables applications to provide context to Large Language Models (LLMs) in a standard way, separating the responsibility of providing context for LLMs from the actual LLM interaction.

Learn more: https://modelcontextprotocol.io/introduction

Advantages of MCP

By defining clear specifications for components such as resources (data exposure), prompts (reusable templates), tools (operations), and sampling (completion), MCP simplifies the development process and promotes consistency. A key advantage is flexibility; MCP allows developers to switch between different LLM providers more easily without completely restructuring their tool and data integrations. It offers a structured approach aimed at reducing the complexity typically associated with implementing custom tools for different models. While libraries like Google Agent Development Kit, LangGraph, OpenAI Agents, or Pydantic-AI can facilitate agent construction, MCP specifically focuses on standardizing the interface between agent reasoning (LLM) and its capabilities (tools and data), aiming to create a more interoperable ecosystem.

Project Overview

This repository demonstrates how to use the MCP protocol in conjunction with several popular LLM Agent frameworks, such as Google ADK, LangGraph, OpenAI Agents, and Pydantic-AI. By providing context data and tools, MCP allows developers to interact with different LLM providers in a unified manner.

🚀 Quick Start

Steps:

1. Clone this repository.
2. Install the required packages:

   uv sync
   

3. Set your environment variables in the .env file:

   LOGFIRE_TOKEN=your_logfire_token
   GEMINI_API_KEY=your_gemini_api_key
   OPENAI_API_KEY=your_openai_api_key
   

4. Run any of the following example scripts to see the simple usage of the MCP server:

• Google Agent Development Kit (ADK)
  • basic_mcp_use/adk_mcp.py
• LangGraph Agents
  • basic_mcp_use/langgraph_mcp.py
• OpenAI Agents
  • basic_mcp_use/oai-agent_mcp.py
• Pydantic-AI Agents
  • basic_mcp_use/pydantic_mcp.py

🔧 Technical Details

MCP Architecture

The MCP architecture defines the following main components:

1. Resources: Data and information provided to the LLM.
2. Prompts: Templates used to guide the LLM to generate responses.
3. Tools: Operations or functions that the LLM can call.
4. Sampling: Defines how to obtain completions from the LLM.

Through these components, MCP provides a standardized way to integrate and manage the context information of LLMs.

About Logfire

Logfire is an observability platform developed by the Pydantic team, aiming to simplify the monitoring of AI applications. Its main features include:

• A simple and powerful dashboard
• Python-oriented insights, including rich display of Python objects
• SQL-based querying of application data
• OpenTelemetry support to leverage existing tools
• Pydantic integration for validation analysis

Logfire provides developers with insights into the operation of their code, which is particularly important for understanding the behavior of LLM models.