coros-workout-mcp
Enables Claude to design strength workouts and push them directly to COROS watches via the Training Hub API. It supports searching an exercise catalog, creating workout routines, and listing existing workouts.
README
coros-workout-mcp
MCP server for creating COROS strength workouts via the Training Hub API. Lets Claude design workouts and push them directly to your COROS watch.
See the MCP in action: YouTube walkthrough
Disclaimer
This is an unofficial, community-driven project. It is not affiliated with, endorsed by, or connected to COROS in any way.
This server communicates with the COROS Training Hub using a reverse-engineered, undocumented API that may change or break without notice. Use it at your own risk.
COROS is a trademark of COROS Wearables, Inc. This project is provided as-is with no warranty — see LICENSE for details.
Setup
cd coros-workout-mcp
npm install
npm run build
Usage with Claude Code
claude mcp add coros-workout -- node /path/to/coros-workout-mcp/dist/src/index.js
To use env var auth (avoids typing credentials in conversation):
claude mcp add coros-workout -e COROS_EMAIL=you@example.com -e COROS_PASSWORD=yourpass -e COROS_REGION=eu -- node /path/to/coros-workout-mcp/dist/src/index.js
Usage with Claude Desktop
Add to ~/Library/Application Support/Claude/claude_desktop_config.json:
{
"mcpServers": {
"coros-workout": {
"command": "/path/to/node",
"args": ["/path/to/coros-workout-mcp/dist/src/index.js"],
"env": {
"COROS_EMAIL": "you@example.com",
"COROS_PASSWORD": "yourpass",
"COROS_REGION": "eu"
}
}
}
}
Node.js 18+ required — this server uses native
fetch()which was added in Node 18.
Troubleshooting
fetch is not defined: Claude Desktop is a GUI app that doesn't inherit your shell PATH, so node binaries installed via version managers (mise, nvm, fnm, volta) won't be found. Use the full absolute path to your node binary in"command":which node # e.g. /Users/you/.mise/shims/node node --version # confirm it's 18+Common locations:
- mise:
~/.local/share/mise/installs/node/<version>/bin/node- nvm:
~/.nvm/versions/node/<version>/bin/node- fnm:
~/.local/share/fnm/node-versions/<version>/installation/bin/node- Homebrew:
/opt/homebrew/bin/node
Tools
| Tool | Description |
|---|---|
authenticate_coros |
Log in with email/password (or auto-login from env vars) |
check_coros_auth |
Verify current auth status |
search_exercises |
Search ~383 exercises by name, muscle, body part, equipment |
create_workout |
Build and push a strength workout to COROS |
update_exercises |
Fetch the latest exercise catalog from COROS and rebuild locally |
list_workouts |
List existing workouts |
Example conversation
"Search for chest exercises with bodyweight"
"Create a workout called 'Quick Push' with 4x15 Push-ups, 3x10 Diamond Push-ups, and 3x20 Decline Push-ups with 45s rest"
Updating the exercise catalog
The bundled exercise catalog (data/exercises.json) is a static snapshot. If COROS adds new exercises, use the update_exercises tool to refresh it. This fetches the latest exercises from the COROS API and i18n strings from the CDN, rebuilds the catalog, and reloads the in-memory cache — all in a single tool call. Requires authentication.
Auth notes
- Region:
eu(Europe) orus(US). Defaults toeu. - Session conflict: Logging in via this API invalidates your COROS web app session, and vice versa.
- Auth tokens are stored at
~/.config/coros-workout-mcp/auth.json(mode 0600).
Development
npm test # Run unit tests
npm run test:watch # Watch mode
npm run build # Compile TypeScript
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。