An AI life agent · early access

The AI that calls you when you don't reply.

You've seen AI coding agents take a goal and work at it until it's done. LifeLoop is that, pointed at your life instead of your codebase. It plans your day, chats with you about it, and phones you when a notification isn't landing. You just talk — the AI understands, rebuilds the plan around you, and remembers everything that actually happened.

Android first. One AI agent across your goals, your habits and your health — not seven apps that don't talk to each other.

LifeLoop is calling you 22:15 · voice

LifeLoop

Hey — your 10pm tablet is still unconfirmed, and it's the antibiotic. Did you take it?

You

not yet, still out at dinner
No problem. I'll check back at eleven. You've got the 6am run tomorrow — want me to move it to seven?

You

yeah do that
Done — run's at 7am, and I'll ask about the tablet at 11. Enjoy dinner.
Understood, re-planned and logged — no forms call · 24s

A conversation, not an app

It talks. You talk back.

No forms, no dropdowns, no logging screens to fill in. You tell LifeLoop what happened the way you'd tell a person, and the AI turns it into structure behind the scenes. It reaches you three ways, and escalates through them when you go quiet.

It calls you

A real phone call when messages aren't landing. You speak, it listens and answers — “I'm in a meeting, give me twenty” is a complete answer.

It chats with you

Type it, or send a voice note. “Skipped the gym, took the tablet after lunch, still feel rough” becomes four entries and one good follow-up question.

It nudges you

When a tap is enough, it stays a tap. The AI decides which channel a moment deserves instead of firing the same alarm at everything.

The problem

Your apps are all waiting for you to open them.

The todo list, the habit tracker, the medicine reminder, the fitness app. Every one of them is passive, and the day you most need them is the day you don't open a single one.

Nothing follows up

A notification fires once and dies. It never notices you swiped it away, never asks again, never escalates when it matters.

Nothing adapts

One meeting runs over and the rest of your plan is fiction. You're the one who has to rebuild the day, so usually nobody does.

Nothing remembers

Six months on you can't say when the goal stalled, which weeks you drank more, or how long that illness lasted. The record was never kept.

How it works

The AI plans, calls, adapts, and remembers.

That's the loop, and it's the whole product. An assistant that only does the first step is a calendar. One that does all four is something you can actually rely on.

01

Plan

The AI builds your day around your goals, your calendar and what you said matters this week.

02

Check in

Notification, then message, then an actual phone call. You talk back in your own words.

03

Adapt

“I'm in a meeting, remind me in 20” is a valid answer. The rest of the day reshapes around it.

04

Remember

What got done, what slipped, and what was going on around it — kept as a record you own.

What it kept from that day Mon, 14 Jul
06:40
Run, 5 km done week 3 of half-marathon plan
08:05
Azithromycin 500 mg, after food taken
11:00
Deep work block — moved to 15:00, your 11:30 ran over
19:20
You logged: drinks with the team
22:00
Evening dose no reply
22:15
AI called — “give me 10 minutes” · re-asked at 22:25
Recorded, correctable, and yours 14 days logged

Because it remembers, it can

  • Show you that you slip most on the nights you drink
  • Notice a goal has been sliding for three weeks, not three days
  • Connect what you took to how you felt afterwards
  • Tell your doctor a story you'd never reconstruct from memory

Because it's proactive, it will

  • Interrupt you — on purpose, at the moment it matters
  • Call when a notification isn't enough
  • Reach your emergency contact if you go quiet on something serious
  • Always let you say “not now” without losing the thread

What it looks after

One agent, not seven apps.

Each area has its own long-term memory, and they talk to each other — which is the entire point. Your habit tracker has never once known you were on antibiotics.

01

Your goals and your day

The big thing you said you wanted, broken into things that actually happen this week — and rebuilt when the week doesn't cooperate.

  • Daily plan that knows your calendar and your energy
  • Re-plans instead of guilt-tripping when you fall behind
  • Long-term progress, not just today's checkboxes

02

Habits you're trying to change

Drinking, screen time, whatever yours is. You log the urge, not just the slip — because the urge you resisted is the part worth knowing about.

  • Captures the context: time, mood, who you were with
  • Finds your actual pattern instead of scoring you
  • Never shaming — a broken streak isn't a lost month
  • The private things sit behind their own lock

03

Health and medication

The area with the least room for error, so it gets the most care. Doses, symptoms, side effects and recovery, kept as they happen.

  • Every dose marked taken, late, missed, skipped or unconfirmed
  • Illness episodes with a symptom timeline
  • Stock estimates and a refill nudge before you run out
  • A doctor-ready summary to take to your appointment

04

The things you'd want found

Blood group, allergies, current medicines and an emergency contact — reachable from the lock screen, offline, by someone who isn't you.

  • Works with no signal, because that's when it's needed
  • Shows medical facts without exposing your identity
  • Separate from everything private, always

On an ordinary Tuesday

The stuff you'd ask a friend, if you had one who remembered everything.

Not features — actual messages. This is what an AI life agent is for: the small, unglamorous decisions you make every day and get no help with.

Eating for a goal

i want to put on 5kg but i'm veg and i hate cooking
Then let's not build a plan that assumes you'll cook. Here's a week at ~2,700 kcal from food you already eat — four of the seven dinners need no stove. I'll adjust the numbers as your weight actually moves.

It plans around the person you are, not the one a diet app wishes you were — and corrects itself from real results.

Is it actually working?

📷two photos · 6 weeks apart
is this cream doing anything or am i imagining it
Same angle, same light — that's why I nagged you about it. Redness along the jaw is clearly down. The spot on your left cheek looks unchanged. Six weeks in, that's real but partial. Want me to keep the comparison going, or is it time to ask a dermatologist?

It compares your own photos over time, so “I think it's helping” becomes something you can actually look at.

Getting dressed

📷mirror photo
wedding at 7. is this ok
Fit's good. The shirt and the shoes are fighting — swap to the brown pair and you're done. Also it's 34°C at 7pm, so the linen one if you own it.

Small confidence questions nobody has anyone to ask, answered in ten seconds.

When the plan is too hard

You've missed the 5am run four days running. I don't think that's discipline — I think it's a bad plan. Want me to move it to 7am and cut it to 3km until it sticks? We can build back up once it's boring.
yeah honestly 5am was never happening

It notices the plan is the problem before you quit — and it's the AI that brings it up, not you.

When you slip

drank again last night. 12 day streak gone, i'm done with this
Twelve days still happened — that doesn't un-count. And look at the shape of it: you were at fourteen nights a month in April, you're at four now. That's the number that matters, not the streak. Same plan tonight?

The moment most apps lose people. No shame, no zeroed counter — just the long-term trend you can't see from inside a bad night.

When you're about to

craving hitting hard rn
You logged the urge instead of acting on it — that's the whole game, and it's the fourth time this month you've done it. It's 11:40pm and you're home alone, which is when this usually lands. Two minutes of breathing, or want me to call Arun?

You log the urge, not just the slip. Over weeks that becomes a pattern you can finally do something about.

Where we're careful

The model runs the conversation. It doesn't decide the medicine.

Language models are excellent at planning a day, asking a good follow-up question and turning a messy week into a record. They are the wrong thing to trust for a drug interaction. So in the places where being confidently wrong actually hurts someone, we don't let the model be the authority.

01

Normalise

Resolve what you typed into a standard drug identity — brand, ingredient, strength, form.

02

Look up

Query a maintained medication knowledge source, not the open internet and not model memory.

03

Apply rules

Deterministic, versioned, testable checks — duplicates, interactions, allergy conflicts.

04

Explain

The model's only job here: say the result in plain language, with its source attached.

Potentially important interaction

These medicines have a documented interaction that may increase a specific risk. Do not stop either medicine based only on this alert. Contact your doctor or pharmacist to confirm what you should do.

Source: [source] · Last checked: [date] · View supporting information

Every important alert names its source, shows when it was last checked, and links out to it. It never says “these cannot be taken together” unless the source clearly says so. The export is a doctor-ready summary, not a doctor-verified report — it labels what you entered and what the AI organised, and it becomes verified only when a professional reviews it.

Privacy

This ends up holding more about you than anything else you own.

Where you were, what you're struggling with, what you take, things you haven't told anyone. Privacy is a design constraint here, not a policy page written the week before launch.

Encrypted in transit and at rest

With strong authentication and sessions that expire on their own.

Consent is specific

Agreeing to reminders is not agreeing to phone calls. Each is asked for separately.

A vault for the private things

Locked behind its own PIN, never in an export, never on the emergency screen.

Export and delete, always

Your record leaves with you, and deletion means deletion.

Never sold, never trained on

No sale of personal information. No model training on your identifiable data by default.

Access is logged

Tightly controlled internal access, recorded, with a documented breach-response process.

Where we are

Early, and honest about it.

We're building the loop first — plan, check in, adapt, remember — because everything else is decoration if the assistant can't reliably reach you and keep an accurate record.

Now · building

The loop, and the memory under it

Daily planning, escalating check-ins from notification to call, medication and habit logging, illness episodes, the doctor-ready export, and the privacy, correction and deletion controls underneath all of it.

Next

Cross-domain intelligence

Areas that negotiate with each other, patterns worth surfacing, and the source-grounded safety layer — normalised medicines, a maintained interaction database, allergy checks and clinician review of how every alert is worded.

Later

Everything it should already know

Prescription and lab-report scanning, wearables, richer voice, caregiver access, and a way for a doctor to review and correct a record directly.

Questions

The things people ask first.

What is an AI life agent?

An AI life agent is an assistant that works at a goal on its own rather than waiting for you to open an app. LifeLoop applies the idea to your life instead of a codebase: it plans your day, checks in on it, adapts when the day changes, and keeps the record. Plan, check in, adapt, remember — that loop is the whole product.

Does LifeLoop actually call you on the phone?

Yes. LifeLoop reaches you three ways — a notification, a chat message, then a real phone call — and escalates through them when you go quiet. On a call you just speak: “I'm in a meeting, give me twenty” is a complete answer, and the rest of the day reshapes around it.

How is LifeLoop different from a to-do list or habit tracker?

A to-do list, habit tracker, medicine reminder and fitness app are all passive — they wait for you to open them, a notification fires once and dies, and none of them remember enough to tell you what changed six months ago. LifeLoop is proactive, and it is one agent across all of those areas, so your habit tracking knows you are on antibiotics.

Can LifeLoop track medication?

Yes. Every dose is marked taken, late, missed, skipped or unconfirmed, illness episodes are kept as a symptom timeline, stock estimates trigger a refill nudge before you run out, and you can export a doctor-ready summary for an appointment.

Is LifeLoop safe to rely on for medical questions?

LifeLoop supports you, it does not diagnose you. Where being confidently wrong would hurt someone, the language model is not the authority: medicines are normalised to a standard identity, checked against a maintained medication knowledge source, and run through deterministic, versioned rules. The model's only job there is to explain the result in plain language with its source attached. It does not replace a doctor, pharmacist or emergency service.

What happens to my data?

Data is encrypted in transit and at rest, consent is asked for separately per channel (agreeing to reminders is not agreeing to phone calls), private entries sit behind their own PIN-locked vault and never appear in an export or on the emergency screen, and you can export or delete your record at any time. Personal information is never sold and identifiable data is not used to train models by default. The full detail is in our privacy policy.

Which platforms is LifeLoop available on?

Android first. LifeLoop is in early access and people are being let in a small group at a time — join the waitlist and you'll get one email when it's your turn.

Waitlist

Be there when it starts calling.

We're letting people in a small group at a time, starting on Android. Leave your email and we'll tell you when it's your turn — nothing else.

One email when we're ready for you. No newsletter, no sharing your address, and you can ask us to delete it at any time.

Careers

We're hiring one product development intern.

One role, one person, joining early enough to shape what this becomes. You'd take a piece of the product from rough idea to something people are using — spec it, design it, build it, put it in front of real users, then decide what happens next.

Think you'd be good at this?

Tell us what you've built, what you'd want to own here, and one thing on this page you think we've got wrong. That last one counts for more than a CV.