tekweld-ecommerce-mcp
MCP server that wraps the Tekweld ecommerce test API, providing tools to browse the mega menu and make generic API calls.
README
tekweld-ecommerce-mcp
MCP server that wraps the Tekweld ecommerce test API
(https://ecommercetest.tekweld.com/api/v1).
Tools
- get_mega_menu — raw payload from
GET /home/get_mega_menu(nested category/collection tree). - list_menu_items — same data, flattened into a simple list with a readable path (e.g.
Men > Shirts > Casual), link, login-required flag, and whether it has sub-items. - call_ecommerce_api — generic passthrough for any other endpoint under the same base URL. Pass
path,method, and optionalquery/body/headers.
Setup
cd tekweld-ecommerce-mcp
npm install
npm run build
This produces build/index.js, a stdio MCP server.
Run standalone (for testing)
npm start
It will print tekweld-ecommerce-mcp server running on stdio to stderr and wait for
MCP JSON-RPC messages on stdin.
Connect to Claude Desktop
Add to your claude_desktop_config.json:
{
"mcpServers": {
"tekweld-ecommerce": {
"command": "node",
"args": ["/absolute/path/to/tekweld-ecommerce-mcp/build/index.js"]
}
}
}
Restart Claude Desktop, then ask it to use the get_mega_menu or list_menu_items tool.
Connect to Claude Code
claude mcp add tekweld-ecommerce -- node /absolute/path/to/tekweld-ecommerce-mcp/build/index.js
Run claude mcp list afterward to confirm it registered, then start a new
claude session (existing sessions won't pick it up) and check /mcp.
Deploying as a remote server (for Cowork / claude.ai custom connectors)
Cowork and claude.ai can only reach remote MCP servers (a public HTTPS URL) —
they cannot see a local stdio server on your machine, even one registered in
claude_desktop_config.json. To use these tools from Cowork or claude.ai, deploy
the HTTP entrypoint (src/http.ts / build/http.js) somewhere public. Render's
free tier works well for this:
- Push this folder to a GitHub repo (Render deploys from a repo, not a local folder).
- On render.com, New → Blueprint, point it at your repo.
It will read
render.yamlin this folder and configure everything automatically (build command, start command, health check).- Alternatively, without the blueprint: New → Web Service → connect repo →
Build Command
npm install && npm run build→ Start Commandnpm run start:http.
- Alternatively, without the blueprint: New → Web Service → connect repo →
Build Command
- Deploy. Render gives you a URL like
https://tekweld-ecommerce-mcp.onrender.com. - Confirm it's up:
curl https://<your-app>.onrender.com/healthshould return{"status":"ok",...}. - In Claude, go to Customize > Connectors (claude.ai) → "+" → "Add custom
connector" → enter
https://<your-app>.onrender.com/mcpas the server URL. - Enable it for a conversation via the "+" button → Connectors, then ask Claude
to use
get_mega_menuorlist_menu_items.
Note: Render's free tier spins the service down after ~15 minutes of inactivity. The first request after a gap will be slow (30-60s) while it wakes back up — that's normal, not a bug.
Run the HTTP server locally (optional, for testing before deploy)
npm run build
npm run start:http
This starts an Express server on $PORT (defaults to 3000) with:
POST /mcp— the MCP Streamable HTTP endpoint (stateless mode)GET /health— health check
Extending
The API has more endpoints under /home/..., /product/..., etc. Rather than
hardcoding each one, use call_ecommerce_api to hit them directly, or add a new
server.registerTool(...) block in src/tools.ts (shared by both the stdio and
HTTP entrypoints) following the pattern used for get_mega_menu.
If the API later requires an auth token, pass it via the headers argument of
call_ecommerce_api, or add an Authorization header default in fetchJson()
in src/tools.ts.
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。