Whoop MCP Server
Enables Claude to read your Whoop recovery, sleep, cycle, and workout data through natural language queries.
README
Whoop MCP Server
A small personal server that lets Claude read your Whoop recovery, sleep, cycle, and workout data. Once it's running, you can ask Claude things like "pull my recovery for this week" and it will fetch real numbers from Whoop.
It does not need to know about your menstrual cycle dashboard — you stay in the loop: ask Claude to pull your Whoop data, then ask it to write a summary into your dossier artifact.
How it works
- You log in to Whoop once, in your browser, directly on this server.
- The server stores your Whoop tokens and refreshes them automatically.
- Claude talks to this server over one endpoint:
/mcp. - This server talks to Whoop's real API on your behalf.
You ──(one-time login)──> this server ──(OAuth)──> Whoop
Claude ──(tool calls)──> this server ──(API calls)──> Whoop
1. Create a Whoop developer app
- Go to https://developer-dashboard.whoop.com and sign in with your Whoop account.
- Create a Team (if prompted), then create an App.
- Note your Client ID and Client Secret.
- Add a Redirect URL. It must be:
https://YOUR-DEPLOYED-URL/auth/whoop/callback(You'll know your real URL after step 3 below — you can come back and edit this.) - Request these scopes:
read:recovery,read:cycles,read:sleep,read:workout,read:profile,read:body_measurement,offline.
2. Configure
cp .env.example .env
Fill in WHOOP_CLIENT_ID, WHOOP_CLIENT_SECRET, and WHOOP_REDIRECT_URI
(this last one must exactly match what you registered in step 1, including
https:// and no trailing slash).
3. Run it somewhere reachable on the internet
Claude connects to this server from Anthropic's cloud, not from your laptop — so it needs a real public HTTPS URL. Any standard Node.js host works (Render, Railway, Fly.io, a small VPS, etc.):
npm install
npm start
Set the same environment variables from .env in your host's dashboard.
The app listens on process.env.PORT (most hosts set this for you
automatically).
About storage: by default, your Whoop tokens are saved to a local file
(./data/whoop-tokens.json). If your host wipes its disk between restarts
(common on some free tiers), you'll need to redo the one-time login after
every restart — or set UPSTASH_REDIS_REST_URL / UPSTASH_REDIS_REST_TOKEN
(Upstash has a free tier) so tokens are stored durably instead. No code
changes needed, just those two environment variables.
Once deployed, go back to your Whoop app settings and make sure the Redirect URL matches your live URL exactly.
4. Connect your Whoop account (one time)
Visit https://YOUR-DEPLOYED-URL/auth/whoop/start in a browser, log in to
Whoop, and approve access. You'll see a confirmation page when it's done.
5. Add this server to Claude
- In Claude, go to Settings → Connectors.
- Click Add custom connector.
- Paste:
https://YOUR-DEPLOYED-URL/mcp - Click Add, then enable it for your conversation (the "+" button → Connectors).
You can now ask Claude things like:
- "What was my recovery this morning?"
- "Pull my last 7 days of sleep data."
- "Compare my strain over the last 2 weeks."
Security, honestly
This server doesn't have its own login wall on the /mcp endpoint — it's
protected only by the URL being private to you. Treat the URL like a
password:
- Don't share it or post it anywhere public.
- Don't commit
.envto a public GitHub repo (.gitignorealready excludes it). - If you want stronger protection later (e.g. real per-request auth), that's a reasonable next step — just ask and we can add it.
Troubleshooting
If something breaks — a deploy error, a 401 from Whoop, Claude saying the connector won't connect — copy the exact error message back to Claude and we'll fix it together. This code was written without the ability to run or deploy it directly, so the first run-through is where real-world issues (dependency versions, host quirks) tend to surface.
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。