402 MCP Restaurant Finder
Enables AI agents to browse restaurants and menus for free, and place orders with USDC payment via x402.
README
402 MCP — Restaurant Finder with Paid Tools
A demo MCP server that shows agentic commerce with x402: AI agents can browse restaurants and menus for free, but must pay USDC to place an order.
Built on Cloudflare Workers + Durable Objects using the Agents SDK.
Tools
| Tool | Cost | Description |
|---|---|---|
list_restaurants |
Free | List all restaurants with cuisine and rating |
show_menu |
Free | Show the menu for a restaurant by ID |
add_to_cart |
Free | Add an item to the cart |
pay_order |
0.05 USDC | Place an order (requires x402 payment) |
Free tools return 200 OK. The pay_order tool returns 402 Payment Required with machine-readable payment instructions until payment is settled.
Prerequisites
- Node.js 20+ and wrangler (
npm i -g wrangler) - A wallet — generate an EVM private key:
npx viem generate-private-key - Testnet USDC — get free USDC on Base Sepolia from the Circle Faucet
- Cloudflare account —
wrangler login
Setup
Install dependencies:
npm install
Set your wallet address as the payment recipient in src/index.ts:
const X402_CONFIG: X402Config = {
network: "base-sepolia",
recipient: "0xYourWalletAddress", // ← replace with your wallet address
facilitator: { url: "https://x402.org/facilitator" },
};
Run locally
npm run dev
The MCP server is available at http://localhost:8787/mcp.
Deploy
npm run deploy
The MCP server will be available at https://402-mcp.<your-subdomain>.workers.dev/mcp.
Test the 402 response
Call the paid pay_order tool without payment — it returns 402:
curl -X POST http://localhost:8787/mcp \
-H "Content-Type: application/json" \
-d '{
"jsonrpc": "2.0",
"id": 1,
"method": "tools/call",
"params": { "name": "pay_order", "arguments": { "order_id": 1 } }
}'
Free tools like list_restaurants return 200 with no payment required.
Connect from an MCP client
Cloudflare AI Playground
- Go to https://playground.ai.cloudflare.com/
- Enter your MCP server URL (
https://402-mcp.<your-subdomain>.workers.dev/mcp) - Use the tools directly from the playground
Claude Desktop
Use the mcp-remote proxy. In Claude Desktop, go to Settings > Developer > Edit Config:
{
"mcpServers": {
"restaurant-finder": {
"command": "npx",
"args": [
"mcp-remote",
"http://localhost:8787/mcp"
]
}
}
}
Restart Claude and the tools will be available.
How it works
The server uses withX402() from agents/x402 to wrap the McpServer. Free tools are registered with this.server.registerTool(...). Paid tools use this.server.paidTool(...) with a price in USDC — the x402 middleware intercepts calls and returns 402 Payment Required with payment instructions until the client settles payment on-chain.
export class MyMCP extends McpAgent {
server = withX402(
new McpServer({ name: "Restaurant Finder", version: "1.0.0" }),
X402_CONFIG,
);
async init() {
// Free tool
this.server.registerTool("list_restaurants", { ... }, async () => { ... });
// Paid tool — 0.05 USDC per call
this.server.paidTool("pay_order", "Pay to place order", 0.05, { ... }, {}, async () => { ... });
}
}
References
| Resource | URL |
|---|---|
| x402 docs | https://developers.cloudflare.com/agents/x402 |
| Charge for MCP tools | https://developers.cloudflare.com/agents/tools/payments/x402/charge-for-mcp-tools |
| Pay from Agents SDK | https://developers.cloudflare.com/agents/tools/payments/x402/pay-from-agents-sdk |
| x402 protocol spec | https://x402.org |
| Circle Faucet (testnet USDC) | https://faucet.circle.com |
| x402 Playground | https://playground.x402.cloudflare.com |
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。