A daily verdict on what work is costing you. Every morning unwynd scores your bodily capacity against your own baselines, prices today's calendar against it, and suggests one concrete move — shorten a meeting, protect a gap. The score is fully deterministic: published weights, auditable ledger, no AI deciding your number.
Timing and metadata, never content. Calendar events arrive as times only — meeting titles are never stored on our servers. Tasks you connect (Todoist, GitHub, Google Tasks) keep their titles because they're shown back to you — but bodies, comments and labels are never read. Sleep and recovery numbers stay on the deterministic scoring path: your health data never reaches an AI model. That wall is structural, not a policy — the code path doesn't exist.
The "ask anything" chat answers from your own event store. Deterministic questions ("how many meetings this week") never touch a model at all. When a model is used, it receives metadata summaries — never health scalars, never meal text, never mood notes. Answers are verified against computed facts before they reach you; when the data is too thin, unwynd says so instead of guessing.
No. Every calendar change is preview-before-save: unwynd shows you exactly what would change, you confirm, and a receipt with a 24-hour undo lands in your record. Automation levels are earned, never default.
Google Tasks is built and switches on as soon as its Google-side registration completes. Oura's reading client is part-built but not connectable yet, so we don't list it. Linear, Jira, Asana, ClickUp, TickTick and Calendly are defined in our connector framework but each needs a developer-app registration with that vendor before we can offer it. Structured, Tiimo and Griply publish no public API — we can't connect what a vendor doesn't expose, and we won't pretend otherwise.
Set a per-day work window in Sources → Work hours — the day unwynd draws and prices follows it. Fully rotating shift patterns (changing week to week) aren't supported yet; that's on the record, not hidden.
The capacity score needs about 5 nights of sleep data to build your baselines. Until then it tells you exactly how many nights remain instead of inventing a number. Learned effects (what a late meeting does to your sleep) build over the first weeks.
Yes, in one tap each: full export (everything we hold, machine-readable) and full delete. Delete means delete — rows gone, tokens revoked. No retention tricks, no "30-day recovery" holding your data hostage.
unwynd · how we protect you · privacy · terms