- uv启动mcp服务的python脚本
uv run --with mcp[cli] mcp run path/to/mcp.py- uvx启动mcp服务的python脚本
uvx --with pillow --with mcp[cli] fastmcp run path/to/mcp.py# sse 模式
http://127.0.0.1:8000/sse # 默认8000端口
http://host.docker.internal:8000/sse # docker容器内访问
# streamable-http 模式
http://127.0.0.1:8000/mcp # 默认8000端口
http://host.docker.internal:8000/mcp # docker容器内访问// 涉及3种模式：stdio，sse，streamable_http
{
// streamable_http模式
"mcpServers": {
"math": {
"transport": "streamable_http",
"url": "http://127.0.0.1:8000/mcp"
},
// sse模式
"weather": {
"transport": "sse",
"url": "http://127.0.0.1:8000/sse"
},
"12306-mcp": {
"type": "sse",
"url": "https://mcp.api-inference.modelscope.cn/sse/76ee2dbba8d74f"
},
"sequentialthinking": {
"headers": {
"Accept": "application/json, text/event-stream",
"Authorization": "Bearer 0e51a8cd-e77f-072ec3f3d161"
},
"url": "https://mcp.api-inference.modelscope.cn/sse/bb19488b5bc049"
},
// 本地服务 stdio模式
"quickchart-server": {
"isActive": true,
"command": "cmd",
"args": [
"/c",
"npx",
"-y",
"@gongrzhe/quickchart-mcp-server"
],
"name": "quickchart-server"
},
"excel-mcp-server": {
"isActive": true,
"command": "cmd",
"args": [
"/c",
"npx",
"--yes",
"@zhiweixu/excel-mcp-server"
],
"env": {
"LOG_PATH": "C:\\Users\\username\\Desktop\\temp",
"CACHE_MAX_AGE": "1",
"CACHE_CLEANUP_INTERVAL": "4",
"LOG_RETENTION_DAYS": "7",
"LOG_CLEANUP_INTERVAL": "24"
},
"name": "excel-mcp-server"
}
}
}如果需要在客户端以 stdio 模式启动本项目：
- 环境准备： 在项目根目录下运行以下命令，安装 MCP 命令行工具：
uv add "mcp[cli]"- 启动配置：
在配置客户端时，使用
uv作为命令，并配合--directory参数指定项目路径。- 命令:
uv - 参数:
["--directory", "/Users/hfhfn/charmProjects/mcp_server/", "run", "main.py"]
- 命令: