MCP 服务开发到发布
MCP 服务开发到发布
环境
MCP服务开发环境需要Python 3.10+
我推荐使用uv工具管理Python环境 https://docs.astral.sh/uv/
这个创建一下标准的包结构, 还是比较方便
MCP 调试工具 : https://github.com/modelcontextprotocol/inspector
MCP 服务
官方文档: https://modelcontextprotocol.io/quickstart/server
MCP 服务
MCP主要通信方式：
- Stdio transport
- Uses standard input/output for communication
- Ideal for local processes
- Streamable HTTP transport
- Uses HTTP with optional Server-Sent Events for streaming
- HTTP POST for client-to-server messages
stdio：通过标准输入输出流传递数据，实现简单，兼容性强，适合本地开发测试。
sse：允许服务器向客户端推送实时更新，单向通信，适用于需要持续数据更新的场景如实时通知。
注:后来阅读文档, SSE的模式已经废弃了, 但还是很多人用
建立项目
mcp_requests % uv init . --package -p 3.13
Initialized project mcp-requests at /Users/xxxxx/Desktop/mcp_requests
mcp_requests % uv add “mcp[cli]”
简单代码, 帮我简单发一个 http 的 get 和post 这种方法
import json
from typing import Dict, Any, Optional
import httpx
from mcp.server.fastmcp import FastMCP# Create an MCP server
mcp = FastMCP("HTTP Requests")@mcp.tool()
def http_get(url: str, headers: Optional[Dict[str, str]] = None) -> str:"""Make an HTTP GET request"""try:with httpx.Client() as client:response = client.get(url, headers=headers or {})return json.dumps({"status_code": response.status_code,"headers": dict(response.headers),"content": response.text}, indent=2)except Exception as e:return f"Error: {str(e)}"@mcp.tool()
def http_post(url: str, data: Optional[Dict[str, Any]] = None, headers: Optional[Dict[str, str]] = None) -> str:"""Make an HTTP POST request"""try:with httpx.Client() as client:response = client.post(url, json=data, headers=headers or {})return json.dumps({"status_code": response.status_code,"headers": dict(response.headers),"content": response.text}, indent=2)except Exception as e:return f"Error: {str(e)}"@mcp.tool()
def http_put(url: str, data: Optional[Dict[str, Any]] = None, headers: Optional[Dict[str, str]] = None) -> str:"""Make an HTTP PUT request"""try:with httpx.Client() as client:response = client.put(url, json=data, headers=headers or {})return json.dumps({"status_code": response.status_code,"headers": dict(response.headers),"content": response.text}, indent=2)except Exception as e:return f"Error: {str(e)}"@mcp.tool()
def http_delete(url: str, headers: Optional[Dict[str, str]] = None) -> str:"""Make an HTTP DELETE request"""try:with httpx.Client() as client:response = client.delete(url, headers=headers or {})return json.dumps({"status_code": response.status_code,"headers": dict(response.headers),"content": response.text}, indent=2)except Exception as e:return f"Error: {str(e)}"def main() -> None:import asyncioasyncio.run(mcp.run(transport='stdio'))if __name__ == "__main__":main()
尝试本地测试调用, 再 cursor (其他也可以,我只是电脑恰好有, 比较方便) 里面添加服务,
{"mcpServers": {"request-server-aaddb": {"name": "request-mcp-server-aaddb","type": "stdio","description": "AADDb MCP server","isActive": true,"command": "uv","args": ["--directory",
"/Users/yi.zhai/Desktop/mcp_requests", "run", "mcp-requests"]}}
}
成功启动
尝试让 AI 调用
服务成功收到请求, 测试成功, 功能是跑通了
PyPI 上传步骤总结
安装构建工具
uv add build twine
构建
uv run python -m build
上传
uv run python -m twine upload dist/* -u **token** -p pypi-xxxxxx
成功标志：
在这里插入图片描述
最后尝试一下 ,直接导入使用, 没问题
{"mcpServers": {"mcp-request": {"command": "mcp-request","type": "stdio"}}
}
最后文档服务发布到了
https://pypi.org/project/mcp-request/ 大家可以试试