大模型应用开发新范式：深入解读MCP协议设计思想 原创

from mcp.server import FastMCP

# 初始化 MCP 服务器，名称用于客户端识别
mcp = FastMCP("my-server")

# 注册工具：同步函数示例（加法器）
@mcp.tool()
def add(a: int, b: int) -> int:
    """计算两个数的和"""
    return a + b

# 注册工具：异步函数示例（模拟API请求）
@mcp.tool()
asyncdef fetch_data(url: str) -> str:
    """从URL获取数据"""
    import httpx
    asyncwith httpx.AsyncClient() as client:
        response = await client.get(url)
        return response.text

if __name__ == "__main__":
    mcp.run(transport="stdio")  # 使用标准输入输出通信

mcp run server.py --transport=sse

from mcp.client.stdio import stdio_client
from mcp import ClientSession, StdioServerParameters, types
import asyncio

# Client会使用这里的配置来启动本地MCP Server
server_params = StdioServerParameters(
    command="python3",
    args=["./mcp_server.py"],
    env=None
)


asyncdef main():
    asyncwith stdio_client(server_params) as (read, write):
        asyncwith ClientSession(
                read, write, sampling_callback=None
        ) as session:
            await session.initialize()
            print(await session.list_tools())
            print('\n正在调用工具...')
            result = await session.call_tool("add", {"a": 1, "b": 2})
            print(result.content)
   time.sleep(60)

asyncio.run(main())

import sqlite3
from contextlib import contextmanager

@mcp.resource()
@contextmanager
def db_connection():
    """数据库连接资源（自动关闭）"""
    conn = sqlite3.connect("data.db")
    try:
        yield conn
    finally:
        conn.close()

@mcp.tool()
def query_users(query: str, conn: sqlite3.Connection = mcp.depends(db_connection)) -> list:
    """执行SQL查询"""
    return conn.execute(query).fetchall()

from fastmcp import FastMCP
import openai  # 官方OpenAI库

mcp = FastMCP("openai-integration")

# 配置OpenAI API密钥（实际应从环境变量读取）
openai.api_key = "sk-your-api-key"

@mcp.prompt()
def role_based_prompt(role: str):
    """动态Prompt：根据角色设定AI行为"""
    role_profiles = {
        "teacher": "你是一个严谨的数学老师，必须逐步解释推导过程",
        "joker": "用幽默的段子和网络流行语回答问题",
        "coder": "仅返回代码，不要任何解释"
    }
    return {"system": role_profiles.get(role, "默认助手模式")}

# 给工具绑定prompt
@mcp.tool(prompt=role_based_prompt)
def ask_question(question: str, role: str = "teacher") -> str:
    """实际调用OpenAI生成回答（非模拟）"""
    response = openai.ChatCompletion.create(
        model="gpt-3.5-turbo",
        messages=[
            {"role": "system", "content": role_based_prompt(role)["system"]},
            {"role": "user", "content": question}
        ],
        temperature=0.7if role == "joker"else0.3# 幽默模式需要更高随机性
    )
    return response.choices[0].message["content"]  # 返回AI生成的文本

if __name__ == "__main__":
    mcp.run(transport="http", port=8000)  # 启动HTTP服务