whoop

Name: whoop
Author: shashankswe2020-ux

Local

github.com/shashankswe2020-uxHealth175Updated 3d ago

Repository

About whoop

whoop is a read-only MCP server that lets users explore WHOOP health and training records through compatible assistants. It provides personal baselines, trends, summaries, and race-performance analysis, with local tools for race logs, workout corrections, and a dashboard.

Key features

01Read recovery, sleep, and workout data
02Compare trends and personal baselines
03Generate health review summaries
04Analyze HYROX and triathlon performance
05Audit and correct local race data
06Launch a local performance dashboard

Use cases

01Review recovery and sleep for the week
02Compare HRV with a personal baseline
03Track HYROX or triathlon training
04Inspect and correct logged race data

MCP support

Exposes read-only WHOOP data tools, resources, and prompt templates to MCP clients. Local stdio mode also enables race logging, workout corrections, and the dashboard; remote connections require a deployment and compatible connector.

Configuration

Local server config is the same for every client. Copy it into your client’s MCP config file.

mcp.json · Universal config

{
  "mcpServers": {
    "shashankswe2020-ux-whoop": {
      "command": "npx",
      "args": [
        "-y",
        "whoop-ai-mcp"
      ],
      "env": {
        "WHOOP_CLIENT_ID": "<WHOOP_CLIENT_ID>",
        "WHOOP_CLIENT_SECRET": "<WHOOP_CLIENT_SECRET>"
      }
    }
  }
}

Run commandnpx -y whoop-ai-mcp

Environment variables to set

WHOOP_CLIENT_IDSecretYour WHOOP Developer App Client ID from developer.whoop.com

WHOOP_CLIENT_SECRETSecretYour WHOOP Developer App Client Secret from developer.whoop.com

Local servers need a runtime on your machine (e.g. Node.js 18+). Config file locations differ by client; see your client’s docs.

Packages

npmwhoop-ai-mcp0.5.2stdio

Tools

No tool list yet

Local servers expose tools only once running on your machine, and the maintainer has not provided a tool list.

Details

From the project README

whoop-ai-mcp is a read-only WHOOP MCP server for Claude, Gemini CLI, Cursor, VS Code, OpenClaw, Perplexity on macOS, and other MCP clients. Ask about your recovery, compare weeks of sleep, or check today against your own baseline, without exporting the data yourself. ChatGPT, Claude web/mobile, and the Grok xAI API require a remote deployment; see the client setup guide.

npx -y whoop-ai-mcp@latest setup --client=claude-desktop

Quickstart | Watch the walkthrough | Website | Tool reference

WHOOP MCP in Claude Desktop: a weekly health review with recovery, HRV, sleep and workout summaries
_{Claude Desktop builds the review; whoop-ai-mcp supplies the data. Your results and layout will differ.}

Ask your WHOOP data anything

Ask	You get
"How am I doing today?"	Recovery, primary sleep, strain and your latest workout, with gaps flagged where data is missing
"Is my HRV unusual for me?"	Your personal baseline: distribution, percentiles and how many days it's based on
"How did my sleep change this month?"	Trends, period-to-period comparisons, sleep debt and consistency
"Give me my weekly health review."	Recovery, sleep and training in one summary

Statistics describe your recorded history. They aren't medical diagnoses or predictions.

Quickstart

You need: Node.js 20 or later, an active WHOOP membership, and a WHOOP Developer App with the redirect URL http://localhost:3000/callback.

Set up. Run the command for your assistant and enter your developer app's client ID and secret when asked.

Assistant	Command
Claude Desktop	`npx -y whoop-ai-mcp@latest setup --client=claude-desktop`
Claude Code	`npx -y whoop-ai-mcp@latest setup --client=claude-code`
Codex	`npx -y whoop-ai-mcp@latest setup --client=codex`
GitHub Copilot	`npx -y whoop-ai-mcp@latest setup --client=copilot`
Gemini CLI	`npx -y whoop-ai-mcp@latest setup --client=gemini-cli`
Cursor	`npx -y whoop-ai-mcp@latest setup --client=cursor`
VS Code	`npx -y whoop-ai-mcp@latest setup --client=vscode`
OpenClaw	`npx -y whoop-ai-mcp@latest setup --client=openclaw`
Perplexity (macOS app + helper)	`npx -y whoop-ai-mcp@latest setup --client=perplexity`

For Claude Code, Codex and VS Code/Copilot, run the registration command that setup prints. For Gemini CLI, Cursor and OpenClaw, merge the printed config; for Perplexity, paste the printed launch command into its local connector. Nothing is installed globally, and optional usage telemetry defaults to No.

These targets, including the claude alias for Desktop, require 1.1.0 or later. For a source build, use node dist/index.js setup --client=<target>.

Authorize. Fully quit and reopen your assistant. The first time the server starts, WHOOP opens in your browser so you can approve access.
Ask. Try "Summarize my sleep and recovery this week."

Need manual configuration, environment variables or a remote deployment? See the full setup guide.

Claude web/mobile and ChatGPT quick setup

Run setup --client=claude-web or setup --client=chatgpt with either implementation to print remote deployment and connector instructions. These are guidance-only targets, not one-click hosting: provide a public HTTPS /mcp endpoint, complete WHOOP authorization on the host, and configure compatible connector OAuth. Availability depends on the client account/workspace. setup --client=grok prints an xAI API MCP tool example; consumer Grok chat integration is not verified. No remote credentials are collected and no client connection is tested.

Prefer a native binary?

The Rust implementation offers the same client targets and limitations without a Node.js runtime for the server. It needs Rust 1.88 or later:

cargo install whoop-mcp
whoop-mcp setup --client=claude-desktop

The additional Rust targets require 0.2.0 or later. For a source build, use cargo run -- setup --client=<target> in the Rust checkout.

How it works

Your WHOOP account grants read-only access to whoop-ai-mcp, which runs on your computer with 16 read-only tools and no hosted relay. It sends tool results to your AI assistant: Claude Desktop, Claude Code, Codex or GitHub Copilot.

Read-only. The server has no tools that create, change or delete anything in your WHOOP account. Race results and workout corrections are saved only to local files, with an audit log.
Local by default. The server runs on your machine. No hosted relay is involved.
Your assistant sees the results. Health data reaches the assistant provider you use, the same as anything else you share in that chat.
Aggregate mode limits the server to five summary tools and hides per-record data. It reduces what's shared, but it isn't anonymization.
Telemetry is off unless you opt in. It never includes health data or chat text.

Details: data handling and telemetry controls.

What's included

In standard mode, the server provides 16 read-only tools, 4 resources and 5 prompt templates, plus 9 HYROX and triathlon tools. Support for resources, prompts and remote connections depends on your assistant.

HYROX and triathlon tracking

Ask "How did yesterday's HYROX sim go?" or "Am I ready for my triathlon in two weeks?" The race tools:

Rebuild a HYROX session from WHOOP's back-to-back workouts: segment timeline, transition time, time in zone and rule-based insights.
Track weekly running and HYROX load, swim/bike/run balance and bricks, plus logged races with station and leg splits and PRs.
Give a readiness band from recovery, training form, specificity, sleep debt and taper. Every component shows its evidence; there's no opaque score.
Let you audit the numbers: suspect workouts are flagged, you can relabel or exclude workouts, every change is logged, and verify_claim recomputes a number your assistant told you.
Open a chart-first local dashboard (open_performance_dashboard) on 127.0.0.1 with Today, Session, Training, HYROX, Triathlon, Readiness and Audit views. Dark/light themes, local fonts, and review-before-save controls are included.

Chart-first Session view: nine recorded run and erg segments, an average and peak heart-rate profile, run-by-run build and recorded-segment ring. Synthetic sample data.

Readiness view: a five-part evidence ring, Not ready verdict, component bands, rationale and evidence disclosures. Synthetic sample data.

Audit view: flagged workout tiles, rule-based findings and correction controls. Flags never automatically exclude workouts. Synthetic sample data.

_{Screenshots use synthetic sample data.}

New in 1.0.2: the chart-first dashboard

Ask your assistant to "Open my performance dashboard" to launch it locally. The design from the latest main-branch redesign is now part of the working dashboard, not just a prototype. Dark is the default, with a Light/Dark toggle that preserves unsaved forms. Session remains the default view.

Today: recovery ring and five weeks of recovery, HRV, resting HR and sleep stages on one timeline with a shared crosshair. Missing days remain gaps; historical bands describe your visible data, not medical norms.
Training: WHOOP zone minutes by UTC week, including empty weeks, workout strain/duration bubbles and fitness/fatigue/form charts. Local corrections apply.
Session and Readiness: recorded-segment and five-component rings, summary HR charts and evidence. No invented continuous HR traces or HYROX station splits.
Review before saving. Inspect workout corrections and race results, then edit, cancel or confirm. Failed saves preserve your entries.
Keep your context. Filters survive switching views; last-7-day and last-28-day presets use completed UTC days.
Refresh honestly. Refresh and retry controls report when a view loaded, not when WHOOP synced. Reads may use cached data.

Dark chart-first Today dashboard: recovery ring and shared-timeline recovery, HRV, resting heart rate and sleep-stage charts, with gaps and historical bands. Synthetic sample data.
_{Today view with synthetic sample data. WHOOP recovery is not a race-readiness verdict; local workout corrections apply to race views, not this overview.}

Training dashboard showing weekly recorded WHOOP zone minutes, strain and duration workout bubbles, and fitness, fatigue and form. Synthetic sample data.

The race log, workout corrections and dashboard are available only when the server runs locally (stdio). Details: race performance tools and local data.

See all tools and prompts

Group	Tools
Today and baselines	`get_today`, `get_baselines`, `get_sleep_debt`
Trends and reviews	`get_weekly_summary`, `get_trend`, `compare_periods`, `get_calendar`
Records	`get_recovery_collection`, `get_sleep_collection`, `get_workout_collection`, `get_cycle_collection`, `get_sleep_by_id`, `get_workout_by_id`, `get_cycle_by_id`
Profile	`get_profile`, `get_body_measurement`
HYROX and triathlon	`get_hyrox_session_report`, `get_hyrox_performance`, `get_triathlon_performance`, `get_race_readiness`, `audit_performance_data`, `verify_claim`
Local only: race log, corrections, dashboard	`log_race_result`, `annotate_workout`, `open_performance_dashboard`

Prompt templates: weekly_health_review, sleep_analysis, recovery_trend, workout_recap, health_check

Parameters, date expressions, resources and caching are in the tool reference.

Documentation

Guide	What's in it
Installation	Assistants, manual configuration, environment variables and WHOOP authorization
Tool reference	Every tool and its parameters, date expressions, resources, prompts and caching
Privacy	Structured results, aggregate mode and what your assistant provider receives
Telemetry	Consent, opting out, what's collected and how long it's kept
Troubleshooting	Local diagnostics, common errors and MCP Inspector
Deployment	Docker, cloud hosting and remote connector OAuth
Development	Tests, project structure and release workflow
Background	Ecosystem research and the video walkthrough

What changed recently? See the changelog or the latest release.

Feedback and contributing

Something not working, or a question you wish you could ask? Use the buttons above. Each opens a short form that asks for the details needed to help. Issues are public, so never include credentials or health records; security problems go through the security policy instead.

Want to contribute code? Read the contributing guide. Everyone taking part follows the Code of Conduct.

If whoop-ai-mcp is useful to you, a GitHub star helps other people find it, and you can buy me a coffee to support development.

Project growth

_{Downloads include automated installs and may not represent unique users.}

License

MIT. This is an unofficial project and isn't affiliated with WHOOP. A WHOOP membership and API access are separate from this package.

Comments

Markdown supported

New to MCP?

From what the protocol is to installing your first server, these answers cover most first questions.

Read the getting-started guide

An open protocol that defines how AI applications connect to external tools and data in a uniform way. Clients talk to the model; MCP servers provide tools, resources and prompt templates. They exchange standard messages, so one server can be reused by many clients.