hevy-mcp
Unofficial MCP server for Hevy that lets AI assistants design training programs, create routines directly in your Hevy account, and analyze logged workouts.
README
hevy-mcp
Unofficial MCP server for Hevy. Design training programs in chat; your AI creates the routines directly in your Hevy account and analyzes your logged training. Requires a Hevy PRO API key (hevy.com/settings?developer).
Use it
Hosted (easiest — claude.ai web, mobile, desktop)
Add a custom connector in claude.ai → Settings → Connectors:
https://hevy-coach.ajwallacemusic.workers.dev/mcp
You'll be prompted for your Hevy API key; it's stored encrypted in your own OAuth grant and sent only to api.hevyapp.com.
Claude Desktop one-click
Download hevy-mcp.mcpb,
double-click, paste your key when prompted.
Local (Claude Code / Cursor / any stdio client)
claude mcp add hevy -e HEVY_API_KEY=<key> -- npx -y @almostjacked/hevy-mcp
Self-host (free, your own infra)
The published package ships the worker — no clone needed:
mkdir hevy-worker && cd hevy-worker && npm init -y
npm install @almostjacked/hevy-mcp wrangler
cat > wrangler.jsonc <<'EOF'
{
"name": "hevy-mcp",
"main": "node_modules/@almostjacked/hevy-mcp/dist/worker.js",
"compatibility_date": "2026-07-01",
"compatibility_flags": ["nodejs_compat"]
}
EOF
npx wrangler deploy
npx wrangler secret put AUTH_TOKEN # any long random string — becomes your Bearer token
npx wrangler secret put HEVY_API_KEY # from hevy.com/settings?developer
Add your worker URL (https://hevy-mcp.<your-subdomain>.workers.dev/mcp) to clients
that support custom headers — e.g. Claude Code:
claude mcp add --transport http hevy-selfhost <url> --header "Authorization: Bearer <token>".
claude.ai custom connectors don't support custom auth headers; for claude.ai use the
hosted instance above, or front your worker with Cloudflare Access.
Tools
| Tool | Description |
|---|---|
| search_exercises | Search the user's Hevy exercise library by name (fuzzy). Returns template id, exact title, type, muscle group and equipment. |
| list_routine_folders | List the user's routine folders (id + title). Check this before creating a folder — reuse an existing one when the name matches. |
| create_routine_folder | Create a routine folder, or return the existing folder if one with this exact title already exists (never duplicates). |
| list_routines | List the user's existing routines (id, title, folder_id, exercise_count). Use to avoid creating duplicates. |
| create_routine | create a routine (lb-first: weight_lb snaps to 2.5/5 lb increments; RPE folded into notes; supersets; duration/distance targets supported) |
| get_workouts | Get the user's most recent logged workouts, newest first, as compact summaries: title, date, duration, and per-exercise set count + top set. Use get_workout for full set-by-set detail of one workout. |
| get_workout | Get one logged workout with every set (weights in kg). Use the id from get_workouts. |
| get_workout_count | Total number of workouts the user has ever logged in Hevy. |
| get_exercise_history | past sets + estimated 1RM in lb for one exercise (full history, optional date range) |
| get_routine | Get one routine in full (all exercises and sets). Always call this before update_routine so you can send back the complete routine. |
| update_routine | Replace an existing routine. WARNING: this overwrites the routine — Hevy replaces its exercises with exactly what you send, so first call get_routine and include every exercise you want to keep. Folder cannot be changed here; duration/distance/custom-metric set targets round-trip safely. |
| get_training_summary | weekly sessions/volume (lb), sets per muscle group, top-lift 1RM trend |
Privacy Policy
Your Hevy API key and training data pass through only to api.hevyapp.com. The stdio/.mcpb versions run entirely on your machine. The hosted instance stores nothing but your encrypted OAuth grant; disconnecting revokes it. Nothing is logged or shared with third parties.
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。