Before you start
These work in any AI app connected to HealthAPI: Claude on every plan, including Free, ChatGPT on Plus and higher, and Cursor, Claude Code or Mistral Vibe. Each lookup uses one of your daily requests, and syncing from your iPhone never counts.
How something changed over time
The most common question by far. Name the metric and the period, and say so if you want a chart.
“How has my resting heart rate changed over the last 90 days?”
It eased from 58 to 54 bpm, a little over 1 bpm a month, and the last two weeks were your lowest.
How has my resting heart rate changed over the last 90 days?
Chart my HRV for the last 6 months, with a 7-day average line.
Compare my average daily steps for each of the last 12 months.
Show my weight over the last 6 months and the average change per week.
Has my VO2 max gone up or down this year? Give me the monthly averages.
Over the last 60 days, which is trending better: my resting heart rate or my HRV?
One day
Steps, exercise, heart, blood oxygen, workouts and the night's sleep for a single day. Today stays partial until it ends.
Give me a summary of yesterday: steps, exercise minutes, sleep and resting heart rate.
What did my day look like on 3 November?
Compare last Saturday with the three Saturdays before it.
Am I more active today than yesterday so far?
Sleep
Each night belongs to the morning it ended, with deep, core and REM when an Apple Watch recorded them.
How did I sleep last night? Include deep and REM.
Show my last 14 nights as a table: time asleep, deep, REM and bedtime.
How consistent was my bedtime this month? Which nights were more than an hour off?
On nights after a workout, do I sleep longer than on rest days?
A specific moment
Individual readings, for questions a daily number can't answer.
What was my heart rate during my last run, minute by minute?
List my blood pressure readings from the last 30 days, with the time of each.
When was my highest heart rate this week, and was I working out then?
Workouts
Type, start time, minutes, distance and active energy for each workout, plus totals.
List this month's workouts by type, with total time and distance.
How did my last 5 runs compare on distance and pace?
How many hours did I train each week over the last 8 weeks?
A whole week
Monday to Sunday, with a line per day. The current week runs up to today.
Compare this week with last week: steps, exercise minutes and sleep.
Write my weekly review: what improved, what slipped, and one thing to try next week.
Latest readings
The newest reading of each metric, with when it was taken.
What are my latest weight, resting heart rate and VO2 max, and when were they taken?
Checks before you trust an answer
AI apps often start here on their own: about one lookup in eight is this overview.
What data do you have for me, and how up to date is it?
List the numbers you used for that answer, day by day.
Which days in that period have no data?
Is today's number complete, or is the day still going?
Getting better answers
- Name the period. “The last 90 days” is clearer than “lately”, and the AI picks the right lookup for it.
- Ask for the numbers. A table day by day lets you check the answer and spot a day with no data.
- Ask for a chart. Claude and ChatGPT both draw charts from the numbers they read.
- Mind partial days and weeks. Today and the current week are still running, so compare finished ones when it matters.
More on single topics: sleep, training readiness, glucose, a weekly report, your year in review and a summary for your doctor.
Questions
Do these prompts work in ChatGPT and Claude?
Yes, once HealthAPI is connected. Claude connects from its directory on every plan, including Free (steps). ChatGPT connects as a plugin on Plus and higher (steps). Cursor, Claude Code, Mistral Vibe and other AI apps that support MCP work too.
How many of my requests does a prompt use?
Each lookup is one request. Most prompts above take one to three; comparing many separate days takes one per day. The preview includes 25 a day, Pro 100 and Developer 1,000. Syncing from your iPhone never counts.
Why does the AI look at an overview first?
To see which metrics you have and how current they are before it answers, so it doesn't guess. It's one request, and about one lookup in eight is this overview.
Can I paste a screenshot of the Health app instead?
You can, but the AI then estimates numbers from a picture and sees a week at most. Reading the data itself gives exact numbers for any period (why screenshots mislead).
Where do the shares at the top come from?
HealthAPI's server counts each lookup by the name of the tool that was called. It never sees your conversation or your question, only that a lookup such as get_trend ran.
Is this medical advice?
No. The AI can spot patterns in your own numbers, but it can't diagnose anything. Talk to a doctor about anything that worries you.
