I turned applehealth into a CLI

The healthai welcome screen — drag-and-drop your export.xml and ask anything.

pipx install healthai
healthai
# drag-and-drop export.xml, ask anything

Why a CLI?

I already ship a Mac app for Apple Health analysis. It works for most people. But every few weeks I'd hear the same thing from a different angle:

The people asking weren't a huge audience — but they were the ones most likely to actually use their health data, fork the parser, file good bug reports, and tell their friends. The Mac app is the front door. The CLI is for the people who'd rather come in through the terminal.

What it does

Same core as the Mac app, in your shell. You point it at export.xml (the file iPhone produces when you tap Export All Health Data), and you can:

• Ask questions in natural language — "What does my sleep pattern look like over the last 30 days?" The CLI extracts the relevant slice and sends only that to the AI provider.
• Run slash commands for canned analyses: /sleep, /steps, /heartrate, /distance, /weight, /workouts.
• Export to CSV or JSON with /csv and /json — full dumps you can pipe into anything.
• Diagnose broken exports with /diagnose. Apple's export format is fragile and I've seen every kind of corruption.

The privacy model

The export.xml file never leaves your machine. The CLI parses it locally with a streaming XML reader (the file can be 500MB+ — you don't want to load that into memory). When you ask a question, only the summarized slice goes out to the AI provider you configured.

Concretely: if you ask "how was my sleep last week?", the AI sees ~7 numbers, not 18 months of every heart-rate sample. That's the whole point of doing the parsing locally. Cloud "health AI" products upload the entire file, which is both slow and a privacy nightmare for what's arguably the most sensitive dataset on your phone.

Why open source?

Two reasons. First — trust. Telling people "your data stays on your machine" works a lot better when they can read the code. The repo is github.com/krumjahn/applehealth, MIT licensed, no telemetry.

Second — the parser itself is useful. Apple's export.xml schema is undocumented in any practical sense, has changed across iOS versions, and includes some genuinely weird edge cases (workouts with negative duration, sleep stages that overlap, units that silently change between metric and imperial). If you want to build anything on top of Apple Health, having a battle-tested parser to fork from is worth more than the AI layer on top.

Install it

# macOS
brew install pipx
pipx install healthai

# Linux / Windows
python3 -m pip install --user pipx
pipx ensurepath
pipx install healthai

Then:

healthai

You'll get the welcome panel above. Drag export.xml onto the terminal window when prompted (or paste the path). That's the whole setup.

What's next

The Mac app already supports LM Studio and Ollama for fully-offline analysis with no API key. That's coming to the CLI next — currently it ships with OpenAI support. If you'd run this against a local model, drop a thumbs-up on the GitHub issue.

Other things on the list, in rough priority order:

• Plain-text mode (no AI) for people who just want fast slash-command summaries
• JSON output flag on every command so you can pipe to jq
• HRV and respiratory-rate commands (currently only via the open-ended question mode)
• An --export-window flag to pre-filter to a date range before analysis

If you try it

I'd genuinely love feedback — what's missing, what's broken, what slash commands you wish existed. Open an issue on GitHub, or email me. The repo has 0 stars at time of writing; if you find this useful, a star helps it surface for the next person searching for "apple health cli".

And if a GUI is more your thing, the Mac app does all of this with point-and-click — same privacy model, no terminal required.