MCP 教程：将麦当劳服务与 AI Agent 集成

API

tutorial

Author

Tony D

Published

November 2, 2025

什么是模型上下文协议 (MCP)？

模型上下文协议 (Model Context Protocol, MCP) 是一项开放标准，旨在使 AI 模型能够无缝连接外部数据源和工具。它由 Anthropic 开发，充当一个通用接口（类似于 USB 接口），允许开发者为大语言模型（LLM）提供对本地文件、数据库和第三方 API 的标准化、安全访问，而无需为每个特定服务编写自定义集成。

通过使用 MCP，AI Agent 可以“触及”物理世界，执行诸如查看库存、查询营养数据，甚至领取优惠券等任务——正如我们将在本教程中使用麦当劳服务所演示的那样。

设置 MCP

获取密钥

从 https://openrouter.ai 获取 LLM 模型 API Key。
从 https://open.mcd.cn/mcp 获取访问 mcd-mcp 的 MCD_TOKEN。

并将它们保存到 .env 文件中：

Code

OPENROUTER_API_KEY=xxxxxx

MCD_TOKEN=XXXX

在代码中使用 MCP

初始化集成 MCP 工具的聊天

Code

library(ellmer)
library(mcptools)
library(dotenv)
# Load environment variables
load_dot_env(file = ".env")

# Define model 
model_name <- "x-ai/grok-4.1-fast"

Code

# Create an ellmer chat object (using OpenRouter as example)
chat <- chat_openrouter(
  system_prompt = "您是一位麦当劳助手。请简短回答。",
  api_key = Sys.getenv("OPENROUTER_API_KEY"),
  model = model_name,
  echo = "output"
)

# Create dynamic configuration with token from environment
config_data <- list(
  mcpServers = list(
    "mcd-mcp" = list(
      command = "npx",
      args = c(
        "mcp-remote",
        "https://mcp.mcd.cn/mcp-servers/mcd-mcp",
        "--header",
        paste0("Authorization: Bearer ", Sys.getenv("MCD_TOKEN"))
      )
    )
  )
)

config_file <- tempfile(fileext = ".json")
jsonlite::write_json(config_data, config_file, auto_unbox = TRUE)

# Load MCP tools via the temporary configuration file
mcd_tools <- mcp_tools(config = config_file)

# Register the MCP tools with the chat
chat$set_tools(mcd_tools)

集成 MCP 的对话测试

现在你可以提问，AI 将使用麦当劳 MCP 工具：

Code

# Query available coupons
chat$chat("目前有什么优惠券")

您当前可用的麦麦省优惠券（均已领取）：

- 北非蛋风味麦满分特惠组合
- 9.9元大薯条
- 9.9元新派组合
- 北非蛋风味麦满分

Code

# Auto-claim all coupons
chat$chat("帮我领取所有优惠券")

暂无可领取的优惠券，所有可用券已领取！

Code

# Get nutrition information
chat$chat("帮我建议一个1000卡路里的套餐，需要包括有糖可乐。")

**推荐套餐（总热量 ≈1011 kcal）：**

- 麦辣鸡腿汉堡（485 kcal）
- 大薯条（379 kcal，使用9.9元优惠券）
- 可乐中杯（147 kcal）

营养均衡，辣味主食+薯条+甜饮，超值！

初始化集成 MCP 工具的聊天

Code

import os
import asyncio
from dotenv import load_dotenv
from langchain_mcp_adapters.client import MultiServerMCPClient
from langchain_openai import ChatOpenAI
from langgraph.prebuilt import create_react_agent

# Load environment variables
load_dotenv()

# Define model (using OpenRouter)
model_name = "x-ai/grok-4.1-fast"

# Create the LLM with OpenRouter
llm = ChatOpenAI(
    model=model_name,
    openai_api_key=os.getenv("OPENROUTER_API_KEY"),
    openai_api_base="https://openrouter.ai/api/v1",
    temperature=0
)

# Build the mcp-remote command with authorization header
mcd_token = os.getenv("MCD_TOKEN")
mcp_command_args = [
    "mcp-remote",
    "https://mcp.mcd.cn/mcp-servers/mcd-mcp",
    "--header",
    f"Authorization: Bearer {mcd_token}"
]

# Configure the MCP client
mcp_config = {
    "mcd-mcp": {
        "command": "npx",
        "args": mcp_command_args,
        "transport": "stdio"
    }
}

async def run_mcp_agent(query: str):
    """Run an MCP-enabled agent with the given query."""
    # Create client (no context manager in v0.1.0+)
    client = MultiServerMCPClient(mcp_config)
    
    # Get tools from MCP server (await required)
    tools = await client.get_tools()
    
    # Create a ReAct agent with the tools
    agent = create_react_agent(llm, tools)
    
    # Run the agent
    result = await agent.ainvoke({
        "messages": [
            {"role": "system", "content": "You are a helpful assistant that can interact with McDonald's services. Please be concise."},
            {"role": "user", "content": query}
        ]
    })
    
    # Return the final response
    return result["messages"][-1].content

集成 MCP 的对话测试

现在你可以提问，AI 将使用麦当劳 MCP 工具：

Code

# Query available coupons
asyncio.run(run_mcp_agent("What coupons are available right now?"))

Code

# Auto-claim all coupons
asyncio.run(run_mcp_agent("Help me claim all available coupons."))

Code

# Check your coupons
asyncio.run(run_mcp_agent("Show me all my current coupons."))

Code

# Query campaign calendar
asyncio.run(run_mcp_agent("What exciting activities are happening in the next 15 days?"))

Code

# Get nutrition information
asyncio.run(run_mcp_agent("Help me design a 1000-calorie meal that includes a sugared cola and costs no more than 30 yuan."))

在 Agent 中使用 MCP

将 “xxx” 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。

JSON 文件配置

Code

{
  "mcpServers": {
    "mcd-mcp": {
      "type": "http",
      "url": "https://mcp.mcd.cn/mcp-servers/mcd-mcp",
      "headers": {
        "Authorization": "Bearer xxx"
      }
    }
  }
}

用户级配置文件:

~/.claude.json

项目级配置文件:

your_project/.mcp.json

提问: “What coupons are available right now?”

将 “xxx” 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。

JSON 文件配置

Code

{
"mcpServers": {
    "mcd-mcp": {
      "url": "https://mcp.mcd.cn/mcp-servers/mcd-mcp",
      "headers": {
        "Authorization": "Bearer xxx"
      }
    }
  } 
}

用户级配置:

~/.gemini/settings.json

项目级配置:

your_project/.gemini/settings.json

提问: “What coupons are available right now?”

将 “xxx” 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。

JSON 文件配置

Code

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "mcd-mcp": {
      "type": "remote",
      "url": "https://mcp.mcd.cn/mcp-servers/mcd-mcp",
      "enabled": true,
      "headers": {
        "Authorization": "Bearer xxx"
      }
    }
  }
}

用户级配置:

~/.opencode.json

项目级配置:

.config/opencode/opencode.json

提问: “What coupons are available right now?”

将 “xxx” 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。

JSON 文件配置

Code

{
  "mcpServers": {
    "mcd-mcp": {
      "serverUrl": "https://mcp.mcd.cn/mcp-servers/mcd-mcp",
      "headers": {
        "Authorization": "Bearer xxx"
      }
    }
  }
}

用户级配置:

~/.gemini/antigravity/mcp_config.json

项目级配置:

暂不支持

提问: “What coupons are available right now?”

将 “xxx” 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。

JSON 文件配置

Code

{
  "mcpServers": {
    "mcd-mcp": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.mcd.cn/mcp-servers/mcd-mcp",
        "--header",
        "Authorization: Bearer xxx"
      ]
    }
  }
}

用户级配置:

~/user/Library/Application Support/Trae/User/mcp.json

项目级配置:

your_project/.trae/mcp.json

开启 Trae 项目级设置: Setting -> Beta -> Enable project MCP

切换模式为 @Builder 并提问: “What coupons are available right now?”

Coming soon

--- title: "MCP 教程：将麦当劳服务与 AI Agent 集成" author: "Tony D" date: "2025-11-02" categories: [AI, API, tutorial] image: "images/0.png" format: html: code-fold: show code-tools: true code-copy: false execute: warning: false editor: markdown: wrap: 72 --- # 什么是模型上下文协议 (MCP)？ **模型上下文协议 (Model Context Protocol, MCP)** 是一项开放标准，旨在使 AI 模型能够无缝连接外部数据源和工具。它由 Anthropic 开发，充当一个通用接口（类似于 USB 接口），允许开发者为大语言模型（LLM）提供对本地文件、数据库和第三方 API 的标准化、安全访问，而无需为每个特定服务编写自定义集成。通过使用 MCP，AI Agent 可以“触及”物理世界，执行诸如查看库存、查询营养数据，甚至领取优惠券等任务——正如我们将在本教程中使用麦当劳服务所演示的那样。 # 设置 MCP ## 获取密钥 1. 从 https://openrouter.ai 获取 LLM 模型 API Key。 2. 从 https://open.mcd.cn/mcp 获取访问 mcd-mcp 的 `MCD_TOKEN`。并将它们保存到 `.env` 文件中： ```{r} #| eval: false OPENROUTER_API_KEY=xxxxxx MCD_TOKEN=XXXX ``` # 在代码中使用 MCP ::: panel-tabset ## R 代码实现 ### 初始化集成 MCP 工具的聊天 ```{r} library(ellmer) library(mcptools) library(dotenv) # Load environment variables load_dot_env(file = ".env") # Define model model_name <- "x-ai/grok-4.1-fast" ``` ```{r} # Create an ellmer chat object (using OpenRouter as example) chat <- chat_openrouter( system_prompt = "您是一位麦当劳助手。请简短回答。", api_key = Sys.getenv("OPENROUTER_API_KEY"), model = model_name, echo = "output" ) # Create dynamic configuration with token from environment config_data <- list( mcpServers = list( "mcd-mcp" = list( command = "npx", args = c( "mcp-remote", "https://mcp.mcd.cn/mcp-servers/mcd-mcp", "--header", paste0("Authorization: Bearer ", Sys.getenv("MCD_TOKEN")) ) ) ) ) config_file <- tempfile(fileext = ".json") jsonlite::write_json(config_data, config_file, auto_unbox = TRUE) # Load MCP tools via the temporary configuration file mcd_tools <- mcp_tools(config = config_file) # Register the MCP tools with the chat chat$set_tools(mcd_tools) ``` ### 集成 MCP 的对话测试现在你可以提问，AI 将使用麦当劳 MCP 工具： ```{r} # Query available coupons chat$chat("目前有什么优惠券") ``` ```{r} # Auto-claim all coupons chat$chat("帮我领取所有优惠券") ``` ```{r} # Get nutrition information chat$chat("帮我建议一个1000卡路里的套餐，需要包括有糖可乐。") ``` ## Python 代码实现 ### 初始化集成 MCP 工具的聊天 ```{python} #| eval: false import os import asyncio from dotenv import load_dotenv from langchain_mcp_adapters.client import MultiServerMCPClient from langchain_openai import ChatOpenAI from langgraph.prebuilt import create_react_agent # Load environment variables load_dotenv() # Define model (using OpenRouter) model_name = "x-ai/grok-4.1-fast" # Create the LLM with OpenRouter llm = ChatOpenAI( model=model_name, openai_api_key=os.getenv("OPENROUTER_API_KEY"), openai_api_base="https://openrouter.ai/api/v1", temperature=0 ) # Build the mcp-remote command with authorization header mcd_token = os.getenv("MCD_TOKEN") mcp_command_args = [ "mcp-remote", "https://mcp.mcd.cn/mcp-servers/mcd-mcp", "--header", f"Authorization: Bearer {mcd_token}" ] # Configure the MCP client mcp_config = { "mcd-mcp": { "command": "npx", "args": mcp_command_args, "transport": "stdio" } } async def run_mcp_agent(query: str): """Run an MCP-enabled agent with the given query.""" # Create client (no context manager in v0.1.0+) client = MultiServerMCPClient(mcp_config) # Get tools from MCP server (await required) tools = await client.get_tools() # Create a ReAct agent with the tools agent = create_react_agent(llm, tools) # Run the agent result = await agent.ainvoke({ "messages": [ {"role": "system", "content": "You are a helpful assistant that can interact with McDonald's services. Please be concise."}, {"role": "user", "content": query} ] }) # Return the final response return result["messages"][-1].content ``` ### 集成 MCP 的对话测试现在你可以提问，AI 将使用麦当劳 MCP 工具： ```{python} #| eval: false # Query available coupons asyncio.run(run_mcp_agent("What coupons are available right now?")) ``` ```{python} #| eval: false # Auto-claim all coupons asyncio.run(run_mcp_agent("Help me claim all available coupons.")) ``` ```{python} #| eval: false # Check your coupons asyncio.run(run_mcp_agent("Show me all my current coupons.")) ``` ```{python} #| eval: false # Query campaign calendar asyncio.run(run_mcp_agent("What exciting activities are happening in the next 15 days?")) ``` ```{python} #| eval: false # Get nutrition information asyncio.run(run_mcp_agent("Help me design a 1000-calorie meal that includes a sugared cola and costs no more than 30 yuan.")) ``` ::: # 在 Agent 中使用 MCP ::: panel-tabset ## Claude Code CLI 将 "xxx" 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。 ### JSON 文件配置 ```{r} #| eval: false { "mcpServers": { "mcd-mcp": { "type": "http", "url": "https://mcp.mcd.cn/mcp-servers/mcd-mcp", "headers": { "Authorization": "Bearer xxx" } } } } ``` *用户级配置文件*: \~/.claude.json *项目级配置文件*: your_project/.mcp.json ### 提问: "What coupons are available right now?" ![](images/clipboard-3343486679.png){width="700"} ## Gemini CLI 将 "xxx" 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。 ### JSON 文件配置 ```{python} #| eval: false { "mcpServers": { "mcd-mcp": { "url": "https://mcp.mcd.cn/mcp-servers/mcd-mcp", "headers": { "Authorization": "Bearer xxx" } } } } ``` *用户级配置*: \~/.gemini/settings.json *项目级配置*: your_project/.gemini/settings.json ### 提问: "What coupons are available right now?" ![](images/clipboard-1202098417.png){width="700"} ## Opencode CLI 将 "xxx" 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。 ### JSON 文件配置 ```{python} #| eval: false { "$schema": "https://opencode.ai/config.json", "mcp": { "mcd-mcp": { "type": "remote", "url": "https://mcp.mcd.cn/mcp-servers/mcd-mcp", "enabled": true, "headers": { "Authorization": "Bearer xxx" } } } } ``` *用户级配置*: \~/.opencode.json *项目级配置*: .config/opencode/opencode.json ### 提问: "What coupons are available right now?" ![](images/clipboard-4015309059.png) ## Antigravity IDE 将 "xxx" 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。 ### JSON 文件配置 ```{python} #| eval: false { "mcpServers": { "mcd-mcp": { "serverUrl": "https://mcp.mcd.cn/mcp-servers/mcd-mcp", "headers": { "Authorization": "Bearer xxx" } } } } ``` *用户级配置*: \~/.gemini/antigravity/mcp_config.json *项目级配置*: 暂不支持 ### 提问: "What coupons are available right now?" ![](images/clipboard-4292207523.png) ## Trae IDE 将 "xxx" 替换为你的 Token，并将以下设置添加到用户级或项目级配置文件中。 ### JSON 文件配置 ```{python} #| eval: false { "mcpServers": { "mcd-mcp": { "command": "npx", "args": [ "mcp-remote", "https://mcp.mcd.cn/mcp-servers/mcd-mcp", "--header", "Authorization: Bearer xxx" ] } } } ``` *用户级配置*: \~/user/Library/Application Support/Trae/User/mcp.json *项目级配置*: your_project/.trae/mcp.json ### 开启 Trae 项目级设置: Setting -> Beta -> Enable project MCP ### 切换模式为 @Builder 并提问: "What coupons are available right now?" ![](images/clipboard-1218453575.png) ## Cody Buddy Coming soon ## Qoder Coming soon ## Cline VS Code Coming soon :::