Can Claude/ChatGPT/Gemini read my entire Apple Health export?

Not effectively. Apple Health exports can be hundreds of MB to 1GB+ and exceed practical context limits. A local API lets you query only the metric and date range you need and send a compact result to the model.

Is my Apple Health data uploaded to the cloud?

The Mac Health Analyzer API runs locally on 127.0.0.1. What gets shared with an AI model is only what you choose to send (typically a small summary or trend payload).

What's the difference between HealthKit and an export-based local API?

HealthKit access is often constrained and app-specific. A local API can query your historical dataset using endpoints designed for summaries and trends across long time ranges.

How do I secure the local API?

Most routes require an X-Health-Analyzer-Token header. You can find the token file path via GET /status. OpenClaw-specific endpoints are read-only and do not require the token.

Local API + AI

Apple Health + AI: Chat With Your Health Data (Local API)

Turn years of Apple Health exports into targeted queries (sleep, steps, workouts, heart rate) that Claude, OpenAI, or Gemini can analyze-without uploading a 1GB file.

Export → AirDrop → Analyze with Claude, ChatGPT, or any AI.

Prerequisite: to use this API, you need the Mac app installed (Mac Health Analyzer / Health Data Export AI Analyzer). The API runs locally while the app is running.

Download Mac App See setup guide

The core idea: LLMs are great at reasoning, but they’re not built to ingest your entire Apple Health export. A local API lets you fetch only the metric slice you need (date range + aggregation), then send that small result to your model.

Base URL: http://127.0.0.1:8765
Auth header (most routes): X-Health-Analyzer-Token: <token>
Token file path: returned by GET /status

Start here: Apple Health Local API

Pick your integration

Apple Health + Claude

Best for tool-use workflows and iterative analysis.

Apple Health + OpenAI / ChatGPT

Best for function calling + structured outputs.

Apple Health + Gemini

Best for recurring weekly/monthly reports.

OpenClaw integration

Automate daily briefs and health workflows (read-only endpoints).

Why a local API beats uploading an export

Apple Health exports are time-series heavy: heart rate samples, sleep stages, workouts, steps, glucose, and more. That means tons of rows and a file that quickly becomes too big for an LLM context window.

Instead of "upload everything", do:

Ask a question (e.g., "How did my resting HR trend over the last 90 days?")
Call a single API endpoint for the metric + date range
Send the compact JSON to the model for analysis

Useful endpoints (examples)

GET /metrics - available metric types
GET /summary?start=YYYY-MM-DD&end=YYYY-MM-DD - compact overall summary
GET /steps/daily?start=YYYY-MM-DD&end=YYYY-MM-DD
GET /sleep/summary?start=YYYY-MM-DD&end=YYYY-MM-DD
GET /workouts/summary?start=YYYY-MM-DD&end=YYYY-MM-DD
GET /heart-rate/trends?start=YYYY-MM-DD&end=YYYY-MM-DD
GET /blood-glucose/summary?start=YYYY-MM-DD&end=YYYY-MM-DD

Security model

Most routes require:

X-Health-Analyzer-Token: <token>

You can discover the token file path via GET /status.

OpenClaw-specific routes are read-only and do not require the token:

GET /openclaw/status
GET /openclaw/daily-brief?date=YYYY-MM-DD

Compatibility routes

Compatibility endpoints also exist under /health/*, for example:

/health/status
/health/sleep
/health/workouts
/health/blood-glucose

Want to talk to your health data?

Start with the Local API docs, then pick your AI integration.

Local API Docs Apple Health + Claude