What they do all day.
One person's day, filling up with people who work for them. Every screen below is the real interface.
06:45
Some of them call.
Any agent can call you in the app instead of texting — a wake-up, a check-in after the gym. It rings like a real call, and you can talk over it mid-sentence the way you would a person.
Sol
is calling you…
⏰ Wake-up call · 6:45
Your alarm, except it's someone.
Rex
02:34
Talk over them mid-sentence and they stop and listen, the way a person does.
Four agents, four voices — tap a card
Each one is cast from its own personality, not picked off a list.
07:00
Turn any of them into a teacher.
Say what you want to learn. Any agent writes the course — eight to fifteen units, in the language you asked in — then teaches it a few minutes at a time, on a schedule, without being asked.
NourHired as: Course coach“I sit the GMAT in November and data sufficiency is where I keep losing points.”
LucaHired as: Language teacher“My partner's family speaks Spanish at Sunday lunch and I follow about a fifth of it.”
VeraHired as: Chief of staff“My bookkeeper sends me the numbers every month and I nod along. I want to read them properly and know what to ask.”
SolHired as: Morning caller“I sit the AWS certification exam in three months. Give me the first ten minutes of the day on it.”
Two of those four were hired to do something else — a chief of staff and a wake-up call.
Nour
AI agent · Course coach
Hey Hadi — good morning. Unit #6: Hidden Constraints change the game. GMAT questions often slip a constraint into the question itself—"positive integers," "distinct values," "non-zero," "real numbers"—that you have to spot. A statement might look insufficient until you remember that constraint, and suddenly it is sufficient. Example: "Is x positive?" and the statement says "x² = 4." Normally that's insufficient (x could be 2 or −2), but if the question already told you x is a positive integer, then x must be 2—so the statement becomes sufficient.
Here's the move: read the question stem first, underline every constraint, then test the statement with those constraints in mind.
One for you: The question is "What is the value of x?" and you're told x is a positive integer. Statement says "x² = 25." Sufficient or insufficient, and why?
9:41Recorded: unit 6, answered correctly
A lesson, not a reminder to have one. It arrived on its own and ended on one thing it can mark. The answer went in correct — and the record answered back: “not mastered yet — needs 2 more correct answers, at least one on a different day.” No agent marks its own work.
When it comes back
Answer something right and you will not see it again for a day. Then three days, then a week, then three weeks — by which point it has stuck.
A unit rests one day after it is taught, then three, then seven, then twenty-one — each step up earned by a right answer on a different day. A wrong answer sends it to the front of the next lesson, whatever the calendar says.Every step has to be earned — by a right answer on a different day. Twice right in one sitting buys nothing.
Forgetting jumps the queue — the thing you missed leads the next lesson, whatever the calendar says. Never more than three at a time.
A lesson counts against the daily cap you set, and a course set up at one in the morning is not taught at one in the morning. Say you need a break and it stops, with no guilt trip.
08:00
An office that runs itself.
Tell one of them that dinner moved to eight. The others hear it and each acts on its part — the morning ring shifts, the shopping goes on today's list, the evening clears.
Basil briefed your team
Sol · tomorrow's ring moved to 6:50✓
Marta · groceries on today's list✓Vera · evening blocked till 8✓Prefer to approve each hand-off yourself? One switch in Settings.
📌Vera
2mThree things matter today: the budget, Karim at 14:00…
Sol
1hMorning! Coffee first, then the deck ☀️
Rex
3hLeg day — hold 80 or add 2.5?
Noor
1dYou sounded lighter today. What changed?
Marta
2dSunday planning — Thursday is leftovers night
Swipe to pin. They keep working either way.
One list, one thread each. Everyone on it works for you.
21:30
Some things go no further.
Say keep this between us — in any language — and it stays with the agent you told. Anything obviously sensitive is marked without you asking.
Noor
AI agent · Health coach
Kept private to Noor
TodayDone — a page you can hand over: what changed since March, and the dates.
There's one thing you asked me to keep private. Include it?
08:24You tell it what you would tell a doctor and nobody else. Weeks later you ask the same agent for the summary you are taking to the doctor: it writes the rest first, then asks. Notice what it does not do — name the thing while asking, in case someone is reading over your shoulder.
Told one, known by one
Where a padlocked thing does not turn up:
- a briefing sent to another agent
- a round table with the others
- the mind map
- a memory file you share
- a notification on your lock screen
- a task on your dashboard
- a search across your agents
- an outside app you gave access to
Where it does: this chat. And on a call — once you have spoken. It never opens with one.
Not a rule the agent is asked to follow — every one of those is a door in the code that a marked fact cannot fit through.
Two limits, said here rather than found out later. Search your own chats and your own sentence is still there — what is hidden is the memory it became, not the day you typed it. And it over-marks on purpose, so you will see padlocks on things you would have been happy to share. Every mark is on the agent’s memory page, says why it is there, and is one tap from off.
Not selling what you say is a promise. This one you can check: not even your other agents.
Also in there
One memory, not twelve
Tell one agent the flight moved and the rest already know. Anything you marked private stays with the agent you told it to.
What changed
Every real change your agents made — a reminder retimed, a fact remembered, a call booked — with when, and who. Read back out of what they saved, not a diary they keep.
Today
What you have on in the next 24 hours, and what your agents heard you promise. It never invents a date.
Agent scorecards
Whether each agent is earning its place — is it answered, are its calls picked up. Scored from what landed, never from what was said.
Round tables
Two to five agents in one call. Say a name and only they answer.
Connections
ClickUp, Todoist, GitHub, Home Assistant — or any web address of your own. Paste the key once and any agent can look there mid-conversation.
Your agents in your calendar
Subscribe once and their schedule sits next to your real meetings. It only goes one way: no sign-in, no permission screen, nothing of yours coming back.
The mind map
Everything your team has learned, drawn as one picture that grows while you talk.
Recordings Inbox
Send in a voice memo or a meeting recording — it comes back written up, and acted on.
Any language
Write in any language and they answer in it — same script, same dialect, decided message by message. Voice too.
Written offline, sent for you
A message the connection could not carry waits its turn and goes when the line comes back. Once, never twice.
Readable with no signal
Your threads are kept on the device as you read them, so a dead line still opens them.
Today, on paper
Print the day as a sheet — schedule, tasks, ruled lines to redraft on — with a QR beside each task that opens the right agent on it.
They reach the phone in your pocket
Install the Android app and their messages arrive as notifications, capped at what you allowed and never more.
On iPhone, add it to your Home Screen
That is what lets their messages land on your lock screen. One tap from the share sheet, and the app tells you how.
Yours to lock, yours to leave
Face ID or a PIN on the app itself. A daily cap you set and an off switch in every message. Take everything with you in one file, whenever you want.
We charge money because we don't sell you.
There are two ways to run a messenger. One of them is to sell what people say in it. We picked the other one, which is why there is a price.
Your conversations stay yours
They are stored so your agents can remember you, and for nothing else. We do not train on them, and the AI companies that write the replies work under terms that keep your words out of their training.
Kept in Europe, locked to your account
Your conversations sit on servers in Europe. The lock is in the database itself, not an app promising to be careful: your account opens your conversations, and no other account can.
Leaving works
Download the lot whenever you want, in one file. Delete is a real delete: agents, conversations, memories, schedules, files — gone at once.
What we learn from is settings, not sentences
Whether a 6:45 ring beats a 7:00 text is worth knowing; what the ring said is not ours to read. A count is only ever kept if at least 25 different accounts are in it, and none of them points back at a person.
No ads, no data brokers, no “anonymised” export to a partner. The detail is in the privacy policy, written to be read.
All of it starts with one agent.
Free · no card · you're in the moment you confirm your email