Commerce MCP Server
Enables AI agents to browse product catalogs, search products with filters, and initiate checkouts, generating order summaries and checkout URLs.
README
MCP Commerce Server Starter
Clone-and-deploy boilerplate for a commerce MCP server. Live on Vercel in five minutes. Reachable by Claude, ChatGPT, Gemini, Cursor, and every other MCP-compatible client.
Full build guide (with the why behind every line): How to Build an MCP Server in 2026
What you get
- Product catalog Resource — agents read the full catalog before acting
search_productsTool — keyword + category + max-price filterinitiate_checkoutTool — returns order summary + checkout URL- Optional API-key auth —
X-API-Keyheader, toggle withREQUIRE_AUTH=true - Health endpoint at
/health - Vercel-ready — one command deploy, free Hobby tier
Stack: Python · FastMCP · FastAPI · Vercel
Quick start (local)
git clone https://github.com/NewPlanetWW/mcp-commerce-starter
cd mcp-commerce-starter
python -m venv .venv && source .venv/bin/activate
pip install -r requirements.txt
python server.py
# Server running at http://localhost:8000
# MCP endpoint: http://localhost:8000/mcp
# Health check: http://localhost:8000/health
Test the health endpoint:
curl http://localhost:8000/health
# {"status":"ok","server":"Commerce MCP Server","version":"1.0.0"}
Deploy to Vercel (free)
Requires Node 18+ for the Vercel CLI. No GitHub required — the CLI uploads directly.
npm i -g vercel
vercel login
vercel --prod
The CLI prints your production URL. Your MCP endpoint is at:
https://your-project.vercel.app/mcp
Set environment variables in the Vercel dashboard (Settings → Environment Variables):
| Variable | Default | Notes |
|---|---|---|
API_KEY |
dev-secret-key |
Change before going live |
REQUIRE_AUTH |
false |
Set true to enforce the key |
Wire into Claude Desktop
Option A — Local stdio (fast iteration while building):
Edit ~/Library/Application Support/Claude/claude_desktop_config.json on macOS (see full guide for Windows/Linux paths):
{
"mcpServers": {
"commerce-catalog": {
"command": "uvicorn",
"args": ["server:app", "--host", "127.0.0.1", "--port", "8001"],
"env": {
"REQUIRE_AUTH": "false"
},
"cwd": "/path/to/mcp-commerce-starter"
}
}
}
Option B — Remote (after Vercel deploy) using mcp-remote:
{
"mcpServers": {
"commerce-catalog-remote": {
"command": "npx",
"args": [
"-y", "mcp-remote@latest",
"https://your-project.vercel.app/mcp",
"--header", "Authorization: Bearer ${MCP_API_KEY}"
],
"env": { "MCP_API_KEY": "your-secret-key" }
}
}
}
Fully quit and relaunch Claude Desktop. Ask: "What products do you have under $100?" — Claude calls search_products and responds with your catalog.
Customize
Replace the sample products
Edit the PRODUCTS list in server.py. Each product needs: sku, name, price, description, availability, category, image_url.
For real inventory, replace the list with a database call:
# server.py — swap PRODUCTS for a live query
import psycopg2 # or SQLAlchemy, Supabase, etc.
def get_products():
# your DB query here
return [...]
PRODUCTS = get_products()
Add Stripe checkout
Replace the stub in initiate_checkout with a real Stripe session:
import stripe
stripe.api_key = os.getenv("STRIPE_SECRET_KEY")
session = stripe.checkout.Session.create(
line_items=[{"price": price_id, "quantity": quantity}],
mode="payment",
success_url="https://yourstore.com/success",
cancel_url="https://yourstore.com/cancel",
)
return {"success": True, "checkout_url": session.url, ...}
The 5 errors you'll hit (and how to fix them)
Covered in the full guide: 30daypivot.com/agentmall_spoke_mcp
MCP error -32600— initialization order violation422 Unprocessable Entity— Pydantic model vs plain args mismatch405 Method Not Allowed— missing DELETE in CORS allowed methodsstateless_http not set— serverless Vercel requiresstateless_http=TrueModuleNotFoundError: mcp— wrong package name (mcp[cli], notfastmcp)
Project structure
mcp-commerce-starter/
├── server.py # FastMCP app — resource, tools, middleware, FastAPI mount
├── requirements.txt # Pinned dependencies
├── vercel.json # Vercel deployment config
├── mcp.json # MCP server manifest
├── .env.example # Environment variable template
└── README.md
Go deeper
This starter is the code companion to the AgentMall spoke series on 30DayPivot:
- MCP Server Build Guide — the full walkthrough behind this repo
- Agent-Readable Product Data — Schema.org markup so agents find your products without calling the server
- FastAPI Commerce API — REST layer that sits alongside your MCP server
- Free-to-Paid / Stripe Metered Billing — monetize the server you just built
- The AgentMall Roadmap — full picks-and-shovels map of agentic commerce infrastructure
License
MIT
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。