Hevy MCP Pro
A production-ready MCP server for the Hevy fitness API that exposes 27 tools for reading, writing, and analyzing workout data, enabling LLM-driven personal trainer workflows.
README
@cdorneles/hevy-mcp-pro
A production-ready Model Context Protocol server for the Hevy fitness API. Exposes 27 tools for reading, writing, and analyzing your workout data — designed for "personal trainer via LLM" workflows in Claude Desktop, Claude Code, and Cursor.
What sets this apart
Compared to the existing community Hevy MCP server, hevy-mcp-pro adds:
- Local analytics: progression series, personal records (Epley 1RM), volume summaries, frequency stats, and stagnation detection — computed locally from your data, not via the Hevy API.
- Smart fetching: single-exercise analytics use Hevy's
/exercise_historyendpoint (one call, full history) instead of paginating all workouts. - Robust HTTP client: token-bucket throttling, exponential backoff with jitter on 429/5xx, honors
Retry-After, and never logs your API key. - Strict request validation: Zod schemas reject malformed payloads before they hit the Hevy API.
- Convenience tools:
duplicate_workout,routine_from_workout,compare_workouts.
Requirements
- Node.js 20+
- A Hevy PRO subscription (required to obtain an API key)
- Your Hevy API key from hevy.com/settings?p=developer
Quick start
Claude Desktop
Edit ~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows):
{
"mcpServers": {
"hevy": {
"command": "npx",
"args": ["-y", "@cdorneles/hevy-mcp-pro"],
"env": {
"HEVY_API_KEY": "your-key-here"
}
}
}
}
Restart Claude Desktop and you should see "hevy" in the MCP tool picker.
Cursor
Add to ~/.cursor/mcp.json:
{
"mcpServers": {
"hevy": {
"command": "npx",
"args": ["-y", "@cdorneles/hevy-mcp-pro"],
"env": { "HEVY_API_KEY": "your-key-here" }
}
}
}
Claude Code
claude mcp add hevy npx -- -y @cdorneles/hevy-mcp-pro
# then set HEVY_API_KEY in the resulting config
Tools
Reading
| Tool | Purpose |
|---|---|
list_workouts |
Paginate workouts (max pageSize 10 — Hevy API limit) |
get_workout |
Fetch a single workout with all sets |
get_workout_count |
Total workout count for the account |
get_workout_events |
Sync incremental updates/deletes since a timestamp |
list_routines |
Paginate routines |
get_routine |
Fetch a single routine |
list_exercise_templates |
Paginate exercise templates |
get_exercise_template |
Fetch one template |
search_exercise_templates |
Local fuzzy search across all templates (cached) |
get_exercise_history |
Flat history of every set for one exercise |
list_routine_folders |
Paginate folders |
get_routine_folder |
Fetch one folder |
get_user_info |
Authenticated user profile |
Writing
| Tool | Purpose |
|---|---|
create_workout |
Log a new workout |
update_workout |
Modify an existing workout |
create_routine |
Create a routine template |
update_routine |
Modify an existing routine |
create_routine_folder |
Create a new folder |
create_exercise_template |
Create a custom exercise |
Analytics
| Tool | Purpose |
|---|---|
get_exercise_progression |
Time series of best set per session for one exercise |
get_personal_records |
PRs (heaviest weight, best set, top estimated 1RM) |
get_volume_summary |
Volume grouped by exercise / muscle group / week |
get_workout_frequency |
Workouts/week, current streak, longest gap |
detect_stagnation |
Exercises whose 1RM has plateaued in the last N weeks |
Convenience
| Tool | Purpose |
|---|---|
duplicate_workout |
Re-log a past workout starting now (or at a given time) |
routine_from_workout |
Convert a logged workout into a reusable routine |
compare_workouts |
Side-by-side diff of two workouts |
Configuration
All configuration is via environment variables.
| Variable | Required | Default | Description |
|---|---|---|---|
HEVY_API_KEY |
yes | — | Your Hevy API key |
LOG_LEVEL |
no | info |
error | warn | info | debug |
HEVY_RATE_LIMIT_CAPACITY |
no | 10 |
Token bucket capacity (max burst) |
HEVY_RATE_LIMIT_REFILL_PER_SEC |
no | 10 |
Token refill rate (requests/sec) |
HEVY_ANALYTICS_MAX_PAGES |
no | 50 |
Soft cap for analytics pagination (50 pages = 500 workouts) |
HEVY_CACHE_TTL_MS |
no | 3600000 |
Exercise template cache TTL (default 1h) |
Example prompts
Once installed, you can ask your LLM things like:
"Show me my bench press progression over the last 3 months."
"What are my current personal records on squats, deadlifts, and bench?"
"Detect any exercises where I've stagnated in the last 8 weeks."
"Duplicate yesterday's workout and start it now."
"Compare my last two leg days."
"How many times did I train chest in March?"
Troubleshooting
Invalid API key or Hevy PRO subscription required
Confirm you have a Hevy PRO subscription and that the key from hevy.com/settings?p=developer is correctly set in HEVY_API_KEY. The key never appears in logs, so a typo is the most common cause.
Rate limited by Hevy API (retries exhausted)
The default client-side throttle is 10 req/s. If you're running heavy analytics queries against a long workout history, lower HEVY_RATE_LIMIT_REFILL_PER_SEC to e.g. 5 to slow down further.
Analysis limited to first 500 workouts
Analytics tools paginate workouts with a soft cap. Raise HEVY_ANALYTICS_MAX_PAGES if you need to analyze deeper history. Each page is up to 10 workouts.
Logs
This server writes structured JSON logs to stderr (stdout is reserved for the MCP protocol). To debug, set LOG_LEVEL=debug and check your MCP client's server log view.
Development
git clone https://github.com/cbotworks/mcp-hevy-pro.git
cd hevy-mcp-pro
pnpm install
pnpm test # 41+ unit tests, no network
pnpm typecheck
pnpm lint
pnpm build # produces dist/ (ESM + CJS + .d.ts)
To regenerate the vendored OpenAPI snapshot:
pnpm fetch-openapi
Architecture
src/
├── client/ HTTP layer (rate limiter, retry, error mapping, paginator)
├── schemas/ Zod schemas for API requests + responses
├── tools/ Tool implementations grouped by domain
│ └── analytics/ Pure analyzers (progression, PRs, volume, frequency, stagnation)
├── utils/ Logger, errors, TTL cache
├── server.ts MCP server: registers tools, wires stdio transport
└── index.ts CLI entry: parses env, fails fast on missing API key
The HTTP client owns all network I/O. Tools never call fetch directly. Analytics tools are pure functions over data fetched via the client, which makes them trivially fixture-testable.
Contributing
Issues and PRs welcome. Please run pnpm typecheck && pnpm lint && pnpm test before submitting.
License
MIT — see LICENSE.
Acknowledgements
The Hevy team for the public API. The community chrisdoc/hevy-mcp server informed early API exploration but no code was reused.
Recommended Servers
playwright-mcp
A Model Context Protocol server that enables LLMs to interact with web pages through structured accessibility snapshots without requiring vision models or screenshots.
Magic Component Platform (MCP)
An AI-powered tool that generates modern UI components from natural language descriptions, integrating with popular IDEs to streamline UI development workflow.
Audiense Insights MCP Server
Enables interaction with Audiense Insights accounts via the Model Context Protocol, facilitating the extraction and analysis of marketing insights and audience data including demographics, behavior, and influencer engagement.
VeyraX MCP
Single MCP tool to connect all your favorite tools: Gmail, Calendar and 40 more.
graphlit-mcp-server
The Model Context Protocol (MCP) Server enables integration between MCP clients and the Graphlit service. Ingest anything from Slack to Gmail to podcast feeds, in addition to web crawling, into a Graphlit project - and then retrieve relevant contents from the MCP client.
Kagi MCP Server
An MCP server that integrates Kagi search capabilities with Claude AI, enabling Claude to perform real-time web searches when answering questions that require up-to-date information.
E2B
Using MCP to run code via e2b.
Neon Database
MCP server for interacting with Neon Management API and databases
Exa Search
A Model Context Protocol (MCP) server lets AI assistants like Claude use the Exa AI Search API for web searches. This setup allows AI models to get real-time web information in a safe and controlled way.
Qdrant Server
This repository is an example of how to create a MCP server for Qdrant, a vector search engine.