What Aide Sees on Your Mac for Wellness — and What Never Leaves It
Wellness only works if you trust it enough to keep enabled.
That trust question usually arrives in one sentence: “What is Aide actually watching?”
This post answers that plainly — for the Mac Wellness agent introduced in Meet Wellness in Aide: Your Mac Work Rhythm, Without the Noise. It is not a legal document. It is the product story in human terms.
The Short Answer
Wellness reuses activity signals AideAI already collects locally on your Mac — work sessions, idle time, meeting load, context switches — through the same pipeline described in Remember What You Saw, Read, and Worked On.
No new permissions are required for the MVP Wellness rollout.
What Wellness adds is interpretation: meaningful breaks, continuous work streaks, load levels in plain language, optional nudges.
What it does not add is surveillance: no window titles, keystrokes, or message content sent to the model for wellness context.
What Stays On Your Mac
These signals are processed locally to build today’s picture and week rhythm:
| Signal | Why Wellness uses it |
|---|---|
| Work sessions | Active vs. idle stretches through the day |
| Idle / lock / away | Pause continuous-work timers when you step away |
| Meeting load | Back-to-back pressure from calendar-backed meetings |
| Context switches | Rough attention fragmentation (rate, not content) |
| Meaningful breaks | Distinguish real pauses from micro-alt-tabs |
Persisted wellness events and daily snapshots live in local storage on your Mac. Analytics for product improvement uses funnel events without PII — not a transcript of your day.
For long-term facts you explicitly save about habits, see Local-First Personal Memory. Wellness and Personal Memory are related ideas (continuity without dumping chat history) but different systems.
What Can Enter Chat (Sanitized Aggregates Only)
When you open Wellness chat or tap Discuss on a notification, the assistant may receive a sanitized snapshot — not a timeline of everything you did.
Typical fields include:
- load level and score band
- continuous work minutes
- meeting minutes today and back-to-back flags
- estimated deep work minutes
- time since last meaningful break
- optional active recommendation category (stretch, walk, end workday, etc.)
Guardrails baked into the product model:
- Cap ~1200 characters for the wellness context block in chat
- Stale after ~5 minutes — old snapshots are not injected as fresh context
- No window titles, bundle IDs as app names, keystrokes, or message bodies
Category breakdowns use safe aggregate labels — not “you were in app X reading document Y.”
What Wellness Never Claims
The in-app disclaimer matters: Wellness is general lifestyle guidance, not medical advice.
Wellness does not:
- diagnose conditions
- replace clinicians or emergency services
- infer mental health states from idle heuristics
- pretend idle detection is perfect (reading at your desk and AFK can look similar — fine for gentle nudges, wrong for clinical claims)
If you need medical or crisis support, use appropriate professional resources — not a Mac assistant.
Anti-Positioning: Not A Surveillance Dashboard
Many “productivity intelligence” tools implicitly sell visibility: screenshots, keystroke counts, always-on monitoring, manager dashboards.
Wellness is intentionally personal and local-first:
- you choose the runtime toggle
- you cap intervention strength
- you set quiet hours
- you dismiss or snooze without guilt-spiral UX
- chat home explains in words, not a leaderboard
That is closer to a calm mirror than a scoreboard.
FAQ-Style Boundaries
Does Wellness read my Messages or Mail content?
No. Wellness context excludes message content. Mail/Notes integrations elsewhere in AideAI follow their own permission and scope rules — separate from wellness runtime aggregates.
Does Wellness send window titles to OpenAI?
No. Sanitized wellness snapshots exclude window titles and similar identifying surface details.
Can I turn it off completely?
Yes. Disable Enable wellness runtime in the Wellness agent’s Settings tab. Notifications stop with the runtime or when you pause them separately.
What if I am not in the rollout cohort yet?
General Wellness chat still works. Personalized Mac insights simply are not enabled until your account receives the feature flag.
How is this different from Activity Log?
Activity Log is your recall timeline for meetings, screenshots, and work context. Wellness is an interpretation layer on activity patterns for rhythm, breaks, and gentle nudges — with stricter sanitization rules for chat injection.
Why This Design Exists
Students and knowledge workers already live inside scores — grades, streaks, inbox counts, due dates. Wellness was built to reduce noise, not add another judgment machine.
Local observation → sanitized explanation → one optional next step is the whole contract.
For how to read chat home block-by-block, read How to Find Wellness in AideAI and What the Chat Home Shows. For notification actions, read Smart Breaks: Wellness Notifications, Accept, and Discuss.
For pricing, visit Pricing.