Moodbound.
Mood tracking, built for bipolar disorder.
The condition itself can prevent you from noticing your own shifts. Moodbound reads your Apple Health data, detects pattern changes with Bayesian models, and tells you calmly when something is moving.
No. 01 · What it is
A diary is not always enough.
Most mood tracking apps are diaries. You log, you read it back, you interpret it yourself. That works for some people. For bipolar disorder, where the condition itself can prevent you from noticing your own shifts, a diary is not enough.
Moodbound runs probabilistic models against your check-ins and your biometric data to catch what you might miss. Low HRV for three days. Sleep dropping below six hours. Activity cratering. Those patterns mean something, and Moodbound connects them to your mood trajectory automatically.
Who this is for
- People living with bipolar I or II who want earlier warning than they get from memory alone.
- People working with a clinician who want a better record between appointments.
- People who want their data to stay on their phone.
Who this is not for
- This is not a diagnostic tool. It will not tell you whether you have bipolar disorder.
- This is not a replacement for clinical care, therapy, or medication management.
- This is not a crisis service. If you are in crisis, please call a local crisis line or 988 in the US.
No. 02 · How it works
Four steps, running quietly in the background.
-
You check in
A short daily check-in: where you are on a seven-point mood scale, how you slept, energy, anxiety, irritability, medications, triggers, a note if you want. Most days it takes under a minute.
-
It reads your body
With your permission, Moodbound pulls sleep, resting heart rate, heart rate variability, step count, and mindful minutes from Apple Health. You choose sleep-only, full integration, or nothing at all.
-
It looks for shifts
Bayesian safety scoring, change-point detection, Wasserstein drift, a hidden Markov model for latent states. It notices when your recent pattern diverges from your usual one.
-
It tells you, calmly
Four severity tiers, no alarms, no shame language. You get information you can take to your clinician, your partner, or just your own reflection.
No. 03 · Features
What's inside.
Check-ins
- Seven-point mood scale from severe depression to mania.
- Energy, sleep hours, irritability, anxiety.
- Medication adherence with structured tracking.
- Custom trigger factors with intensity ratings.
- Free-form notes.
- Adaptive prompts that suggest what to log next, ranked by how much new information it would give the model.
Apple Health
- Reads: sleep, resting heart rate, HRV (SDNN), step count, mindful minutes.
- Writes: State of Mind (iOS 18+), one mindful session per check-in.
- Granular opt-in. Sleep-only or full integration. You can revoke anytime from the Health app.
Analytics
- Bayesian safety engine with four tiers: none, elevated, high, critical.
- Online change-point detection for episode transitions.
- Conformal prediction intervals on 7-day risk forecasts.
- Wasserstein drift scoring for stability.
- Hidden Markov model over latent states: depressive, stable, elevated, unstable.
- Medication trajectory analysis.
- Trigger attribution ranking.
- Weather correlation (rain, temperature).
Safety planning
Heads up: this section briefly mentions warning signs and crisis planning. Skip to the next section if you need to.
- An editable safety plan: warning signs, coping strategies, emergency steps.
- Support contacts with one-tap calling.
- Risk assessment combines mood, sleep, HRV, heart rate, and activity signals.
- Non-clinical language throughout. No diagnoses, no medical claims.
No. 04 · Privacy
Your data belongs in one place: your phone.
No accounts. No sign-up. No analytics. No telemetry. No cloud sync. Your check-ins, your Apple Health permissions, and the model's output all stay on your device, in encrypted storage.
What leaves your phone
- Weather. Your location at one-kilometre accuracy goes to Open-Meteo, then is discarded.
That's the only thing. Nothing else.
What we don't do
- We don't have a backend. There is no we server-side.
- We don't sell, share, or aggregate your data. There's nothing to sell or share.
- We don't collect crash reports, usage stats, or identifiers.
Full details: Privacy policy. Everything here is verifiable; the source is public.
No. 05 · Questions
Things people have asked.
Is this medical advice?
No. Moodbound is a tracking and pattern-detection tool. It is not a diagnostic device, not a substitute for clinical care, and not a crisis service. If it flags something, treat that as information to bring to a clinician, not a verdict.
Do I need Apple Health to use it?
No. You can use Moodbound as a check-in-only app. You'll get less out of the pattern detection without biometric signals, but mood, sleep logs, medications, and triggers still work.
Does my data leave my phone?
Only for weather lookups: your location at 1 km precision, sent to Open-Meteo. Nothing else.
Is there an AI rephraser?
Not in the App Store version. The open source code includes an optional NVC Rephraser, a recommended AI companion that turns reactive statements into the Nonviolent Communication format (observation, feeling, need, request). If you build from source and supply your own AWS Bedrock credentials, you can enable it. When you use it, the text you type is sent over HTTPS to AWS; no other data leaves your phone.
How do I delete everything?
Uninstalling the app deletes all Moodbound data from your phone. You can also export everything to a JSON file from Settings before you do.
How is this different from Bearable, Daylio, or Apple's built-in mood tracking?
Those are diaries. They're good diaries, and for some people a diary is enough. Moodbound is built around the observation that in bipolar disorder the person having the episode is often the last to notice, so it focuses on automatic pattern detection from biometrics rather than self-report alone.
Is it really free? What's the catch?
Really free. No ads, no in-app purchases, no subscription, no premium tier. The source is MIT-licensed and public. The App Store version is just a signed build so you don't have to compile it yourself.
Will there be an Android version?
Not planned. The app depends on HealthKit, which is iOS-only.
How can I report a bug or ask a question?
Open an issue on GitHub.