AI Analysis

Use AI to find patterns, trends, and anomalies in your Apple Health data. The Mac app supports both cloud and local providers, so you can choose the balance of quality, speed, and privacy that works for you.

The app never uploads your health data to our servers. Data goes directly from your Mac to the AI provider you choose, or stays fully local when you use Ollama or LM Studio.

Supported AI Providers

The current Mac app supports these providers out of the box. If you subscribe to Pro, OpenAI and Anthropic are built in and ready to use with no key setup — open Settings and start analyzing. Everyone can use OpenRouter with their own key (the feature is free; you pay OpenRouter for usage), or run a model locally with Ollama or LM Studio. You can pick your default provider in Settings.

Provider Type Cost Best For
OpenAI Cloud Built in for Pro (no key), or bring your own key Strong general analysis, large context windows
Anthropic Cloud Built in for Pro (no key), or bring your own key Nuanced health reasoning, detailed explanations
OpenRouter Cloud Free feature — you pay OpenRouter with your own key Access to many models (Llama, Mistral, Gemini, etc.) through a single API key
Ollama Local Free Maximum privacy, no data leaves your Mac
LM Studio Local Free User-friendly local inference with a GUI for model management

Setting Up Each Provider

Already a Pro subscriber? OpenAI and Anthropic are built in — open Settings, confirm your default provider, and skip straight to running analysis. No API key needed. The steps below are for bringing your own key.

OpenAI

  1. Go to platform.openai.com/api-keys and create an API key.
  2. In the Mac app, open Settings and select OpenAI as your default provider.
  3. Paste your API key into the field provided.
  4. Choose a model available in the app.

Anthropic (Claude)

  1. Go to console.anthropic.com/settings/keys and create an API key.
  2. In the Mac app, open Settings and select Anthropic as your provider.
  3. Paste your API key.
  4. Choose a model available in the app.

OpenRouter

  1. Go to openrouter.ai/keys and create an API key.
  2. In the Mac app, open Settings and select OpenRouter as your provider.
  3. Paste your API key.
  4. Choose from any model you have access to through OpenRouter.

Ollama (Local, Free)

  1. Install Ollama from ollama.com.
  2. Pull a model such as ollama pull llama3.
  3. Make sure Ollama is running on your Mac.
  4. In the Mac app, open Settings and select Ollama as your provider. The app auto-detects the local Ollama server at http://localhost:11434.
  5. No API key is needed. All processing stays on your machine.

LM Studio (Local, Free)

  1. Download LM Studio from lmstudio.ai.
  2. Open LM Studio and download a model (e.g., Llama 3, Mistral, Phi-3).
  3. Start the local server in LM Studio (it runs on http://localhost:1234 by default).
  4. In the Mac app, open Settings and select LM Studio as your provider.
  5. No API key is needed. All processing stays on your machine.

What AI Analysis Can Find

Once you have a provider configured, you can run analysis on your imported Apple Health data. The AI examines your metrics and produces insights that go far beyond what the Health app shows you.

Trends

The AI identifies upward and downward trends in your metrics over time. For example, it might notice that your resting heart rate has been gradually declining over the past three months, or that your step count drops every weekend.

Anomalies

Unusual data points get flagged automatically. A sudden spike in resting heart rate, an unusually short sleep night, or a gap in workout data can all indicate something worth investigating.

Correlations

The AI looks for relationships between different metrics. Does your sleep quality improve on days you walk more than 8,000 steps? Does your HRV drop after late-night workouts? These cross-metric patterns are difficult to spot manually but straightforward for an AI model.

Suggestions

Based on the patterns it finds, the AI generates actionable recommendations. These might include adjusting your bedtime, modifying your workout schedule, or investigating a metric that has changed unexpectedly.

Reading Your Results

AI insights render as formatted text, not raw output. Markdown tables display as real tables, numbered and bulleted lists are laid out cleanly, and code blocks are shown in a monospaced block — so a response full of pipes and asterisks reads the way it was meant to. Numbers follow your region's formatting (for example, thousands separators and decimal marks match your Mac's locale).

Accuracy: Why Your Numbers Should Match Apple Health

The Mac app's analysis metrics are built to match what you see in the Apple Health app. Totals are de-duplicated across devices, so steps and distance recorded by an iPhone and Apple Watch worn together are not double-counted. Sleep reflects time actually asleep rather than time in bed, and distances recorded in miles are converted correctly instead of being read as kilometers.

If your analysis totals do not match the Health app, make sure you are on the latest Mac release — these accuracy improvements shipped in Mac 1.996. After updating, your daily totals should line up with Apple Health.

Privacy Considerations

Your choice of provider directly affects where your health data is processed.

Provider Type Data Location Privacy Level
Cloud (OpenAI, Anthropic, OpenRouter) Data is sent to the provider's servers for processing Your data is transmitted over an encrypted connection. Review each provider's data retention policy. The Mac app strips personally identifiable information before sending.
Local (Ollama, LM Studio) Data never leaves your Mac Maximum privacy. No network requests are made. All inference runs on your local hardware.

If privacy is your top priority, use Ollama or LM Studio. Your health data stays entirely on your machine and no cloud API is involved.

Example Analysis Output

Here is what a typical AI analysis response looks like when you ask for a health summary:

Health Summary — Last 7 Days
============================

Steps
  Average: 8,432/day (up 12% from prior week)
  Best day: Tuesday (12,104)
  Lowest: Sunday (3,210)

Sleep
  Average duration: 7h 14m
  Bedtime consistency: Fair (std dev: 48 min)
  Trend: Slightly improving since last week

Resting Heart Rate
  Average: 58 bpm (stable, within normal range)
  Anomaly: Thursday spike to 67 bpm — likely correlated
  with short sleep the night before (5h 22m)

Workouts
  4 sessions logged (2 running, 1 strength, 1 cycling)
  Total active calories: 2,180 kcal

Suggestions
  1. Your Sunday step count is consistently low. Consider
     adding a short walk to your weekend routine.
  2. Your Thursday heart rate spike correlates with poor
     sleep. Prioritize 7+ hours on Wednesday nights.
  3. You are on a 3-week upward trend in running distance.
     Consider a deload week to support recovery.

Next Steps