Short answers, written honestly — including where the product does not yet do what you might reasonably expect.
That is the correct behaviour and the most common first impression. Amon plans around your commitments and from your work — so with no calendar connected and nothing on your list, there is nothing to plan.
Two things fix it: Integrations → Your calendars → Connect, and add two or three things you actually owe someone on the Tasks screen. Then press Auto-schedule, or type plan my day.
An illustration, so a brand-new account is not a blank screen. They are not your day, nothing is scheduled around them, and they disappear the moment anything real arrives — a connected calendar, a generated plan, or a block you make yourself.
Google Calendar — Integrations → Connect Google Calendar, if your server has it configured. Otherwise use the .ics address below.
Any calendar, including Google, iCloud and Outlook — find its published
or secret .ics address and paste it into Integrations:
Above the timeline, not on it. A holiday describes a day rather than a time — turning Christmas Day into a 00:00–23:59 appointment would fill your calendar with invented hours — so all-day entries sit in a strip at the top of the day instead.
Holidays are read about four months ahead, so you get warning rather than a surprise.
A public holiday does not stop Amon planning. Plenty of people work through them, and the two biggest holiday publishers disagree about whether a holiday even counts as "busy" — so it is shown, and you decide.
"Annual leave" on your own calendar does stop it. If an all-day entry says you are away — annual leave, out of office, PTO, a day off — Amon puts no new work on that day and tells you why. Anything already on the day stays exactly where it is; nothing is deleted.
Work through these in order:
Three, and then it stops: ten minutes before, at the start, and once more five minutes in if you have not begun. No daily digest, no streak guilt, nothing after that.
From the notification itself you can Start, Push 15m, or Not today — without opening the app.
Every block carries its reason. Tap it, or press Why? — you will get the actual constraint that decided the placement, not a generated summary.
If you disagree, drag it. The plan re-checks live and works around your decision: green fits, amber warns, red is refused because something fixed is already there.
That is the honest answer rather than a failure. The day did not hold them.
Amon budgets a realistic fraction of your free time rather than all of it, tightens that budget further after a short night, and refuses to produce a plan you could not actually follow. Unscheduled work stays on your list; it has not been lost.
No. Imported events are fixed — the planner works around them and will never move one. Anything you pin is treated the same way.
The sleep window is absolute. Nothing may be scheduled into it — not by you, not by the planner, whatever the block is.
Move the window itself in Settings → Sleep & shift. Night shifts are supported: the window wraps midnight and moves wholesale.
When you are free, and nothing else. Not what you are doing, not how busy you are, not your email address, not your surname. Just start times, converted into their own time zone.
Press Preview on any link to see precisely what they see.
It appears on your day as a pinned event — because someone else has now written it in their diary, and a planner that quietly moved it would be worse than one that refused to plan at all.
Either side can cancel. The guest gets a single-use link in their confirmation and needs no account; you can cancel from the Meet screen. Cancelling frees the hour.
No — and it is labelled design study for that reason. Those five people are invented and nothing on that grid touches another human.
Scheduling links are the real thing, and they work today. Cross-organisation free/busy needs write access to other people's calendars, which no integration in this product has.
A booking link says when I am free and invites someone to take a slot. A schedule link says what I am doing, and is read-only — nobody can book through it.
The second is a bigger disclosure, so it has stricter defaults: every schedule link must have an expiry date, the maximum is 90 days, and the default detail level shows only that you are busy.
Everything except the times. Blocks that touch or overlap are merged into one stretch before they leave the server, so the reader cannot count your meetings, work out that a fifteen-minute gap is a separate commitment, or tell a focus block from a one-to-one.
Two back-to-back meetings and one two-hour meeting look identical.
They see the titles, and only the titles. Not the reason a block was placed, not the notes, not the linked task, and never a meeting join link — a forwarded link that let a stranger walk into your call would be a far worse leak than the title.
Choose this one deliberately. The app asks you to confirm it, and it does not ask for the other one.
On our own server, and it can only ever encode a link back into Amon. Handing your URL to a public QR service would mean posting a link that reads your calendar to a third party, and a QR generator that accepts any address is a phishing tool with somebody else's domain on it.
The link stops working immediately, and an expired, revoked or never-existed link all give the same answer — so nobody can use the URL to test whether a link is real.
But revoking is not un-seeing. Anyone who already opened it has already read what it showed. The app says so at the point you revoke, rather than implying otherwise.
Because the answer is the only thing that makes “can I take this on?” worth asking. Amon compares how long work actually takes against what you thought it would take, and budgets your next commitment from the difference.
Until it has at least five finished tasks with a recorded time, it says so and takes your estimates at face value — rather than inventing a correction and presenting it as measurement.
No. Press ▷ on a task if you want the clock to do it, or just say how long it took when you tick something off. Both end up in the same place, and both are your own account of your own work.
Skipping the question costs nothing and is one click. A number given quickly to dismiss a prompt looks exactly like evidence and is not, which is worse than a gap.
No. Past eight hours a timer reads as one left running rather than work done, so it is closed, kept, shown to you — and left out of the arithmetic entirely.
If you notice before stopping it, Discard throws it away and records nothing at all.
Yes. A task's recorded time is not a running total that gets nudged — it is the sum of the sittings behind it, recomputed from them each time. That is why stopping a timer twice cannot inflate it, and why deleting a sitting corrects it.
Yes, where your browser supports it — Chrome and Safari do, Firefox does not. The microphone button only appears if it will actually work, rather than being drawn and then apologising.
Press it, say what you want, and it fills the box and sends it. The speaker button beside it reads replies back; it is off until you turn it on, and it remembers.
No. We receive the finished text — the same string you would have typed.
Being exact about the rest: the speech-to-text is done by your browser, and in Chrome that means the audio goes to Google under Chrome's own privacy terms, not to us and not through our API. Safari does it on the device. We never receive, record or store audio either way.
Nothing. It proposes, and waits.
Ask it to add a task, tick one off, change an urgency, move something into a sprint or publish a schedule link, and it shows you exactly what it will do and a button. Nothing is written until you press it, and the action that publishes something looks different from the ones that do not.
Settings → Two-factor sign-in → Turn on. Scan the code with any authenticator app, confirm with a six-digit code, and save the recovery codes somewhere that is not the phone holding the authenticator.
Each recovery code works once. If you lose both the phone and the codes, nobody can get you back in without shell access to the server.
You. The admin console deliberately cannot return plan content, task titles, event titles or sleep data — not as a policy, but because the API has no route that returns them. An administrator can see counts and account metadata, and disable an account.
There is no third-party tracking of any kind in this product. The usage analytics are first-party and record that something happened, never what it was about.
Settings → Your data & devices — export gives you a single JSON file with everything, and delete removes the account and cascades to all of it.
Almost always the date. The app shows today in your account's time zone — check Settings if you have moved country recently.
If it is still empty, your session is fine: nothing is deleted by signing in elsewhere, and your plan syncs across every device you sign in on.
Use the feedback button in the bottom corner of the app. It goes straight to a queue that gets read, and it carries which screen you were on — which is usually the missing half of a bug report.
Useful things to include: what you expected, what happened, and roughly when. Screenshots help more than descriptions.
This is a private beta. Some of what Part 1 describes is not built — alarms, cross-organisation meeting coordination, travel time from a location, and learned energy patterns are all named above or in the product documentation as not yet real. Nothing here is described as working when it is not.