Automating a Morning Health Brief with Claude Code and Apple Health

A morning health brief is useful when it answers two questions: how did I recover last night, and what does that tell me about today? Most health apps show you numbers. A brief that interprets those numbers saves the cognitive step.

Here’s how to build one using Claude Code and health4ai.

What a Useful Brief Covers

The core of a morning brief:

Last night’s sleep — total hours, stage breakdown (Core/Deep/REM)
HRV reading — today’s value vs your recent baseline
Resting heart rate — elevated or normal
Today’s starting point — what the combination implies about readiness

Optional additions:

Workout logged yesterday (was there training stress to account for?)
Trend direction over the last week (is this an isolated dip or part of a pattern?)

The Tools Required

The brief uses two primary tool calls:

get_daily_snapshot(date="2026-06-19")  # today's date
get_hrv_trend(days=14)                 # 2 weeks of HRV for baseline context

get_daily_snapshot returns everything recorded for the day — steps so far, HRV reading, resting HR, sleep records, and any workouts. get_hrv_trend gives the recent baseline so today’s HRV reading has context.

For a more complete brief, add:

get_sleep(days=1)           # detailed sleep stage breakdown for last night
get_coaching_brief()        # full recovery + training load context

Building the Prompt

In Claude Code, a simple morning brief prompt:

Pull today's health snapshot and my HRV trend for the last 14 days. 
Write a 3-sentence morning brief: sleep quality last night, HRV vs recent baseline, 
and what the combination suggests about training today.

Claude calls the two tools, processes the JSON, and returns something like:

Sleep last night was 7h 12min with solid REM (2:18) and good deep sleep (1:05). HRV this morning is 61ms, which is above your 14-day average of 54ms — a meaningful positive signal. Recovery looks good; if you have intensity planned, this is a fine day for it.

That’s the brief. Three sentences, actionable, grounded in data.

Automating It

To run this automatically each morning, use a LaunchAgent that calls Claude in headless mode:

# Save as ~/scripts/morning-health-brief.sh
#!/bin/bash
DATE=$(date +%Y-%m-%d)
claude --print "Pull my health snapshot for $DATE and my HRV trend for 14 days. Write a 3-sentence morning health brief covering sleep quality, HRV vs recent baseline, and readiness for training today." \
  | curl -s -X POST "https://api.telegram.org/bot${TELEGRAM_BOT_TOKEN}/sendMessage" \
    -d "chat_id=${TELEGRAM_CHAT_ID}" \
    -d "text=$(cat -)" \
    -d "parse_mode=Markdown"

Make it executable:

chmod +x ~/scripts/morning-health-brief.sh

Then create the LaunchAgent plist:

<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE plist PUBLIC "-//Apple//DTD PLIST 1.0//EN" "http://www.apple.com/DTDs/PropertyList-1.0.dtd">
<plist version="1.0">
<dict>
  <key>Label</key>
  <string>com.health4ai.morning-brief</string>
  <key>ProgramArguments</key>
  <array>
    <string>/bin/bash</string>
    <string>/Users/yourname/scripts/morning-health-brief.sh</string>
  </array>
  <key>StartCalendarInterval</key>
  <dict>
    <key>Hour</key>
    <integer>7</integer>
    <key>Minute</key>
    <integer>30</integer>
  </dict>
  <key>EnvironmentVariables</key>
  <dict>
    <key>TELEGRAM_BOT_TOKEN</key>
    <string>your_token</string>
    <key>TELEGRAM_CHAT_ID</key>
    <string>your_chat_id</string>
  </dict>
</dict>
</plist>

Save to ~/Library/LaunchAgents/com.health4ai.morning-brief.plist and load it:

launchctl load ~/Library/LaunchAgents/com.health4ai.morning-brief.plist

The brief arrives at 7:30 AM. Apple Watch’s overnight HRV and sleep data will be in HealthKit and synced to Postgres by then via health4ai’s HKObserverQuery listener.

Timing Considerations

Why 7:30 AM and not midnight? Apple Watch often finalizes sleep data (including stage breakdown) during the morning wakeup sequence. Running the brief too early may miss the last hour of sleep. 7:00-8:00 AM is reliable.

What if I slept in? The brief is keyed to the current date, so get_daily_snapshot(date=today) will show whatever HRV and sleep data is in the database at run time. If you wake up at 9 AM and the brief ran at 7:30, the sleep stage data may be incomplete. You can run it again interactively.

Data freshness: health4ai’s iOS app uses HKObserverQuery, which means sync happens when HealthKit pushes new data — typically within seconds of Apple Watch writing a new reading. By 7:30 AM, overnight HRV and sleep data has been in Postgres for hours.

Extending the Brief

A few additions that work well once the basic version is running:

Add context from yesterday:

Also pull workouts from yesterday (get_workouts(days=1)) and mention any training 
that would explain elevated resting HR or suppressed HRV.

Trend flag:

If my HRV trend (get_hrv_trend(days=7)) is declining for 3+ days in a row, 
add a note that this warrants attention.

Weekly summary on Mondays:

Change the LaunchAgent to also trigger a different prompt on Mondays that includes get_health_summary(days=7) and get_coaching_brief() for a fuller weekly review.

health4ai is free through July. Everyone in the founding batch gets lifetime access at $0.
Download on the App Store →