MCP Server Study

🚀 MCP Learning Project ⚡

This project is a learning project based on the MCP protocol, mainly used to learn how to implement an MCP Server.

The api_server directory contains a set of API service interfaces implemented using the FastAPI library, which are used to simulate the Java back - end services of an existing system. In this project, only the four arithmetic operations of addition, subtraction, multiplication, and division are implemented.

The tests directory contains a unit - testing program for the api_server, which is used to get familiar with the pytest functionality.

The mcp_server directory contains the content of the MCP server, including two implementation methods of MCP.

🚀 Quick Start

Prerequisites

• Python 3.10+

Install Dependencies

pip install fastapi uvicorn pytest

Run the Project

1. Start the API service (in the api_server directory)

cd api_server && python main.py

2. Start the MCP Server (in the mcp_server directory)

cd mcp_server && python remote_server.py

3. Configure and use the cline plugin for testing

✨ Features

• 🧮 A toolset for four arithmetic operations (addition, subtraction, multiplication, and division)
• 📡 Real - time message transmission based on SSE
• ⚡ Support for asynchronous HTTP clients
• 🔒 Strongly - typed input validation

📦 Installation

The installation mainly involves installing Python dependencies. You can use the following command to install the required libraries:

pip install fastapi uvicorn pytest

💻 Usage Examples

Basic Usage

There are two implementation methods in the mcp_server directory:

1. STDIO Method

This is mainly for local calls and local operation. The main implementation is included in the server.py code. You can use the cline plugin to call it. Here is the cfg configuration for cline call:

{
  "mcpServers": {
    "math": {
      "timeout": 60,
      "command": "mcp",
      "args": [
        "run",
        "\\mcp_server\\server.py"
      ],
      "transportType": "stdio",
      "disabled": true
    }
  }
}

STDIO Running Mode

# Enter the api_server directory
python main.py
# Start the API service

# After configuring the cline cfg
# Test the mcp server in cline

2. SSE Method

Use Uvicorn to generate a service and configure the connection using the remote server mode of the cline plugin. The main functionality is included in the remote_server.py. Here is the cfg configuration for cline:

{
  "mcpServers": {
    "calculate": {
      "autoApprove": [
        "calculate_sum",
        "calculate_subtract",
        "calculate_multiply",
        "calculate_divide"
      ],
      "disabled": false,
      "timeout": 60,
      "url": "http://127.0.0.1:8001/sse",
      "transportType": "sse"
    }
  }
}

SSE Running Mode

# Enter the api_server directory
python main.py
# Start the API service

# Enter the mcp_server directory
python remote_server.py
# Start the mcp server

# After configuring the cline cfg
# Test the mcp server in cline

🔧 Technical Details

System Architecture

graph LR
    Client-->|SSE Streaming| MCP_Server["MCP Server (Port 8001)"]
    MCP_Server-->|HTTP RPC| API_Server["API Server (Port 8000)"]
    API_Server-.->|Call Function| FUNC[(Calculation Function)]
    
    subgraph Calculation Service
        API_Server
    end

Technology Stack

• Framework: Starlette
• Server: Uvicorn
• HTTP Client: HTTPX
• Protocol: MCP Protocol