garmin-mcp

garmin-mcp

Remote MCP server for Garmin Connect on Cloudflare Workers (free tier), providing 139 tools and 5 resources covering activities, health, training, workouts, nutrition, challenges, devices, gear, weight, women's health, courses, FIT-file analysis, and user profile, with OAuth 2.1 login flow.

Category
访问服务器

README

garmin-mcp

Remote MCP server for Garmin Connect on Cloudflare Workers (free tier). Full TypeScript port of taxuspt/garmin_mcp: all 138 tools plus a garmin_get passthrough (139 total) and the 5 workout template/reference MCP resources.

Modules: activity management (21), health & wellness (29), training & performance (15), workouts + builders (19), nutrition (14), challenges/devices/gear (18), weight/body/women's health/courses (14), FIT-file analysis (4), user profile (4), passthrough (1).

Auth: the server is an OAuth 2.1 authorization server (via @cloudflare/workers-oauth-provider). Adding it to an MCP client opens a browser page where you enter your Garmin email and password (and an MFA code if your account uses 2FA). The Worker runs Garmin's SSO web-widget login flow server-side — the one login path that isn't bot-blocked from datacenter IPs — obtains a long-lived (~1 year) OAuth1 token, discards your credentials, and stores only that token encrypted in the grant. It then mints short-lived OAuth2 bearers on demand for connectapi.garmin.com. A "paste a token" fallback (via the local script below) is tucked behind a details toggle in case Garmin ever starts blocking server-side login.

Deploy

  1. npm install
  2. npx wrangler kv namespace create OAUTH_KV — paste the id into wrangler.jsonc
  3. npx wrangler deploy

Connect a client

claude mcp add garmin --transport http https://garmin-mcp.<your-subdomain>.workers.dev/mcp

Claude Code opens the authorization page in your browser; follow its two steps. claude.ai web: Settings → Connectors → Add custom connector with the same URL. Claude Desktop and other stdio-only clients: mcp-remote handles the OAuth flow too.

Access is OAuth-only — there is no raw-header path. Every /mcp request is gated on a provider-issued access token, and an access token is minted only by completing the login flow.

Locking to your own account

Set OWNER_GARMIN_IDS in wrangler.jsonc to your Garmin displayName (the value get_user_profile returns) — a comma-separated allowlist. Any login for a Garmin account not on the list is rejected at the authorize step, so strangers can't use your deployment even with their own valid Garmin credentials. Leave it empty to allow any Garmin account.

Deviations from the original (remote/serverless adaptations)

  • download_activity_file returns file content inline (fit as base64, gpx/tcx/csv as text, size-capped) instead of writing to local disk; set_fit_download_dir is a documented no-op.
  • upload_course takes gpx_content (GPX XML text) instead of a local file path.
  • get_activity_fit_data covers session/laps/records + power analysis (NP, VI, power curve, W/kg); the original's Di2 shift, climb-detection, and HRV analytics are not ported.
  • Free-tier subrequest budget (~50/invocation): get_training_load_trend capped at 45 days (was 90), get_power_duration_curve at 15 activities (was 20), get_hrv_trend uses the range endpoint in one call, goal/step/menstrual pagination loops capped.
  • Tools return JSON objects rather than pre-serialized strings; errors surface as MCP tool errors.

Notes

  • The Garmin token is stored encrypted inside the grant in Workers KV; when it expires (~1 year), tool calls start failing — reauthenticate the server in your client to run the flow again.
  • Access tokens are bearer credentials: whoever holds a valid one has your data until it expires, so the security rests on your MCP client keeping its stored token secret.
  • Local dev: npm run dev, then point a client at http://localhost:8787/mcp.

推荐服务器

Baidu Map

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。

官方
精选
JavaScript
Playwright MCP Server

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。

官方
精选
TypeScript
Magic Component Platform (MCP)

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。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

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。

官方
精选
本地
TypeScript
VeyraX

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。

官方
精选
本地
graphlit-mcp-server

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。

官方
精选
TypeScript
Kagi MCP Server

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。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

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。

官方
精选