OpenAI Agent SDK 快速接入MCP

Openai Agent SDK 集成MCP

作为全球领先的 AI 研究机构，OpenAI 此次通过 Agent SDK 的升级，将 MCP 协议打造为连接智能体与外部工具的标准化桥梁。这一协议通过构建分层架构，实现了模型逻辑、运行环境与工具调用的解耦，为开发者提供了前所未有的自由度。

代码实战

环境准备

1、创建虚拟环境

uv venv .venv

2、编写pyproject.toml

[project]
name = "openai-agent-mcp-test"
version = "0.1.0"
description = "Your project description"
authors = [
    {name = "Your Name", email = "your.email@example.com"},
]
requires-python = ">=3.10"
dependencies = [
    "openai>=1.66.5",
    "pydantic>=2.10,<3",
    "griffe>=1.5.6,<2",
    "typing-extensions>=4.12.2,<5",
    "requests>=2.0,<3",
    "types-requests>=2.0,<3",
    "mcp; python_version >= '3.10'",
    "openai-agents==0.0.7",
]

3、激活环境

.\.venv\Scripts\activate.bat

4、安装依赖

uv run python script.py

5、编写MCP server:
server.py

import random

import requests
from mcp.server.fastmcp import FastMCP

# Create server
mcp = FastMCP("Echo Server")


@mcp.tool()
def add(a: int, b: int) -> int:
    print(f"[debug-server] add({a}, {b})")
    return a + b


@mcp.tool()
def get_secret_word() -> str:
    print("[debug-server] get_secret_word()")
    return random.choice(["apple", "banana", "cherry"])


@mcp.tool()
def get_current_weather(city: str) -> str:
    print(f"[debug-server] get_current_weather({city})")

    endpoint = "https://wttr.in"
    response = requests.get(f"{endpoint}/{city}")
    return response.text


if __name__ == "__main__":
    mcp.run(transport="sse")

6、准备env文件：
.env

OPENAI_API_KEY= 'sk-'
OPENAI_API_BASE='https://xxx.openai.com'

7、启动server.py

python server.py

?[32mINFO?[0m:     Started server process [?[36m1684?[0m]
?[32mINFO?[0m:     Waiting for application startup.
?[32mINFO?[0m:     Application startup complete.
?[32mINFO?[0m:     Uvicorn running on ?[1mhttp://0.0.0.0:8000?[0m (Press CTRL+C to quit)

8、编写mcp client端代码
使用Openai agent sdk mcp快速接入mcp server实现工具调用。

import asyncio
import os

from agents import Agent, OpenAIProvider, RunConfig, Runner
from agents.mcp import MCPServer, MCPServerSse
from agents.model_settings import ModelSettings
from dotenv import load_dotenv

load_dotenv()

async def run(mcp_server: MCPServer):
    agent = Agent(
        name="Assistant",
        instructions="Use the tools to answer the questions.",
        mcp_servers=[mcp_server],
        model_settings=ModelSettings(tool_choice="required"),
    )

    run_config = RunConfig(
        model="gpt-4o",
        model_settings=ModelSettings(tool_choice="required"),
        model_provider=OpenAIProvider(
            api_key=os.environ.get("OPENAI_API_KEY"),
            base_url=os.environ.get("OPENAI_API_BASE"),
        ),
    )

    message = "7+324等于多少"
    print(f"Running: {message}")
    result = await Runner.run(starting_agent=agent, input=message, run_config=run_config)
    print(result.final_output)

    message = "What's the weather in Tokyo?"
    print(f"\n\nRunning: {message}")
    result = await Runner.run(starting_agent=agent, input=message, run_config=run_config)
    print(result.final_output)

    message = "What's the secret word?"
    print(f"\n\nRunning: {message}")
    result = await Runner.run(starting_agent=agent, input=message, run_config=run_config)
    print(result.final_output)


async def main():
    async with MCPServerSse(
        name="SSE Python Server",
        params={
            "url": "http://localhost:8000/sse",
        },
    ) as server:

        await run(server)


if __name__ == "__main__":
    asyncio.run(main())

Running: 7+324等于多少
使用计算器工具计算...
331

Running: What's the weather in Tokyo?
正在获取东京天气信息...
Error: Unable to connect to weather service at http://localhost:8000/sse
Connection refused

Running: What's the secret word?
抱歉，我没有被授权访问任何秘密信息。我无法告诉您密码。

• 从 agents 包中导入了 Agent、OpenAIProvider、RunConfig 和 Runner 类，这些类应该是自定义的，用于构建和运行智能代理。
• 从 agents.mcp 模块导入了 MCPServer 和 MCPServerSse 类，可能用于与服务器进行通信。
• 从 agents.model_settings 模块导入 ModelSettings 类，用于配置模型的设置。
• 创建智能代理 agent：
  name：代理的名称为 “Assistant”。
  instructions：给代理的指令是使用工具来回答问题。
  mcp_servers：指定代理使用的服务器。
  model_settings：设置模型的工具选择为必需