1.安裝依賴

conda create -n mcpo python=3.11
conda activate mcpo
pip install mcpo pip install uv

2.隨便從https://github.com/modelcontextprotocol/servers?tab=readme-ov-file

找一個mcp服務使用就行，我這里選的是爬蟲

然后安裝

pip install mcp-server-fetch

3.然后我們運行

uvx mcpo --port 8000 -- uvx mcp-server-fetch

4.此時在瀏覽器打開

http://localhost:8000/docs

就可以看到接口文檔了

然后寫個py腳本測試一下

import requests
import jsondef fetch_webpage(url, max_length=10000, start_index=0, raw=False):"""Fetch content from a URL using the MCP Fetch server.Args:url (str): The URL to fetchmax_length (int): Maximum number of characters to returnstart_index (int): Start content from this character indexraw (bool): Get raw HTML content without markdown conversionReturns:dict: The response from the server containing the fetched content"""try:# Make a POST request to the fetch endpointresponse = requests.post("http://localhost:8000/fetch",json={"url": url,"max_length": max_length,"start_index": start_index,"raw": raw})# Ensure the request was successfulresponse.raise_for_status()# Parse the responsereturn response.json()except Exception as e:return {"error": str(e)}# Example usage
if __name__ == "__main__":# Fetch the specific URL you requestedtarget_url = "https://docs.cline.bot"result = fetch_webpage(target_url)print(result)

over

mcpo出現的意義非常好，節省了我們配置的時間，從此，我們就可以像訪問接口一樣使用mcp服務了

5.通過配置文件啟動多個mcp服務同時

mcpo.json

{"mcpServers": {"fetch": {"command": "uvx","args": ["mcp-server-fetch"]},"time": {"command": "uvx","args": ["mcp-server-time", "--local-timezone=Asia/Shanghai"]}}}

執行下面的命令

uvx mcpo --config mcpo.json --port 8000 --api-key "your-secret-key"

?對應接口文檔如下所示