MCP MultiServer Interoperable Agent2Agent LangGraph AI System

This project demonstrates a real-time multi-tool agent architecture based on LangGraph and the MCP protocol. By decoupling agent orchestration and tool execution, it realizes a modular and scalable AI system. The MCP service supports multiple transport protocols (SSE/STDIO), allows AI agents to dynamically call remote tools, and has cross-language and cloud deployment capabilities.

Developer tools Artificial intelligence chatbots #Real-time Agent #Tool Decoupling #Protocol Interoperability #Module Execution .Python

rating : 2.5 points

downloads : 11.7K

update time : 2025-04-29

Open Site

What is the MCP Server?

The MCP server is an AI tool server based on the Model Context Protocol (MCP), used for dynamically executing tools, retrieving knowledge, and supporting complex multi-layer AI interactions.

How to use the MCP Server?

By installing and configuring the MCP server, you can easily integrate AI tools, resources, and task templates, enabling them to be called by AI clients.

Applicable Scenarios

The MCP server is very suitable for application scenarios that require real-time tool execution, knowledge retrieval, and cross-AI system collaboration.

Main Features

Dynamic Tool Execution

Allows AI models to call external tools, such as running terminal commands or getting weather forecasts.

Resource Retrieval

Supports accessing local files or API responses, such as reading files in the desktop directory.

Multi-layer AI Interaction

Supports AI components as clients and servers to achieve modular complex task processing.

🚀 Project Overview

This project focuses on the Model Console Protocol (MCP), a versatile AI interface that enables seamless interaction between external AI applications and structured data sources or executable functions.

✨ Features

Introduction to MCP

Model Console Protocol (MCP): MCP is a universal AI interface that allows external AI applications (such as Claude Desktop or Cursor) to interact with structured data sources and executable functions.
Design Concept: Similar to the USB - C interface, the MCP server acts as a peripheral that can be connected to multiple AI systems.

Roles of MCP Client and Server

MCP Client (embedded in AI hosts like Claude or Cursor)

Request tools, query resources, and handle prompts.
Serve as a bridge between the AI system and the MCP server.

MCP Server (locally implemented)

Expose tools (e.g., weather API) for dynamic invocation by AI clients.
Provide resources (e.g., API responses, database records).
Process prompts to enable structured user interactions.

MCP Functions and Future Enhancements

Agent Composability: The architecture supports multi - layer agent interactions, allowing AI agents to act as both MCP clients and servers. This enables modular and specialized agents to handle different tasks.
Self - evolving AI through API Registration: Future versions can support dynamic tool discovery, enabling AI clients to register and discover new MCP functions in real - time.
Development and Debugging Support: Use Anthropic's MCP Inspector for interactive testing and debugging without full deployment.

MCP and Containerization

Initialize the project with UV, create a virtual environment, and install dependencies (e.g., MCP CLI).
Use Cursor to index the official MCP documentation and update the project according to Cursor rules.

📦 Installation

Implementing a Simple MCP Server

Implement a simple MCP server in server.py using the MCP Python SDK.
The server should expose a function called "Terminal Tool", allowing users to run terminal commands.

Exposing Resources to the MCP Server

Write code using the MCP Python SDK to expose resources on the MCP server.
The goal is to provide the mcpreadme.md file located in the desktop directory as a resource.

📚 Documentation

Avoiding Redundant Work

MCP is similar to the USB - C interface. The MCP server is like an external device that can be connected to AI (e.g., Claude Desktop) or cloud applications.
Functions only need to be written once and can be used as plugins in multiple hosts. The MCP client is embedded in the MCP host to achieve one - to - one interaction with the MCP server. The MCP client invokes tools, queries resources, and inserts prompts, while the MCP server exposes tools (model control: retrieval, database update, sending), resources (application control: database records, API), and prompts (user control: documents).