MCP Pytest Server

🚀 Guide to Developing MCP Server with Pytest

This guide provides a comprehensive overview of developing an MCP server using Pytest, including installation, service startup, debugging, development suggestions, and a summary of the Python SDK implementation.

🚀 Quick Start

📦 Installation

Install Node.js and npm

curl -o- https://raw.githubusercontent.com/nvm-sh/nvm/v0.39.5/install.sh | bash
nvm install --lts

Install Python 3.x and pip

Most Linux systems come pre - installed with Python 3. If not, install it according to your specific operating system. Make sure pip is installed.

💻 Starting the Service

node ~/workspace/mcp-pytest-server/index.js

Check running processes:

ps aux | grep index.js

Capture network traffic (assuming the port is <port_number>):

sudo tcpdump -i any -s 0 -w mcp_traffic.pcap port <port_number>

Interact using Curl:

curl http://localhost:5000

🔧 Debugging the Service

Interact with the service using cline:

use_pytest-mcp

💡 Development Suggestions

Use a Faster JSON Library

Replace the built - in json module with orjson to speed up parsing and serialization.

import orjson as json

Implement a Scheduling Mechanism

Use a dictionary to map request types and tool names to handler functions.

def handle_list_tools(request):
    # Handling logic
    pass

def handle_record_session_start(args):
    # Handling logic
    pass

# Other tool handling functions...

request_handlers = {
    "list_tools": handle_list_tools,
    "call_tool": {
        "record_session_start": handle_record_session_start,
        # Other tools...
    }
}

def handle_request(request):
    request_type = request["type"]
    handler = request_handlers.get(request_type)

    if handler:
        if request_type == "call_tool":
            tool_name = request["name"]
            tool_handler = handler.get(tool_name)
            if tool_handler:
                tool_handler(request["arguments"])
            else:
                send_response({"type": "error", "code": -32601, "message": f"Unknown tool: {tool_name}"})
        else:
            handler(request)
    else:
        send_response({"type": "error", "code": -32601, "message": f"Unknown request type: {request_type}"})

Concurrent Processing

Explore using asynchronous programming (e.g., asyncio) or multi - threading to handle multiple requests. This requires significant modifications to the server structure.

📚 Summary of Python SDK Implementation

Current Status

• Created the Python SDK package structure in ~/workspace/mcp-pytest-server/python-sdk.
• Implemented basic package files:
  • setup.py contains package configuration.
  • src/mcp/__init__.py contains version information.
• Successfully installed it in development mode using pip install -e ..
• Verified the PYTHONPATH configuration to allow package imports.
• Current running status: Development/Testing (not suitable for production use).

Service - Level Details

• Development Mode: Using pip install -e . allows immediate code changes without reinstallation.
• Source Code Access: Full access to the source code for debugging and development.
• Dependency Management: Managed through setup.py, with direct access to the local development environment.
• Stability: Suitable for testing and development, not for production use.
• Performance: May contain debugging logs and unoptimized code paths.

Remaining Tasks

• Implement the core Python functionality of the MCP client.
• Add pytest integration hooks.
• Create a complete test suite for the Python SDK.
• Publish to PyPI for easier distribution.
• Optimize production deployment.