Mila answers at 11 pm. And knows your kids by name.

An ordinary evening

Anna, Berlin, two kids, 9 and 5. Tuesday, 9:40 pm.

Anna: Bedtime took an hour and a half again. Out of bed three times, water, the toilet, a question about dinosaurs. I shouted in the end.

Mila: An hour and a half at the end of a full day, and you still went back in. That counts. For a five-year-old, every trip out of bed is a check that you're still there. Try one thing tomorrow: announce the last visit. «I'll come once more in five minutes, then the door stays open and I'm in the kitchen.» Then actually come. Would that work in your flat?

Anna: Maybe. He'll call anyway.

Mila: Probably, the first two nights. Answer from the kitchen, don't go in. By the third night most five-year-olds stop testing. If it's still an hour and a half next week, tell me, and we look at the afternoon instead.

When it's serious

You: I can't do this anymore. Some evenings I think the kids would be better off without me.

Mila: I'm glad you wrote that here instead of keeping it in. This is bigger than tired, and you don't have to carry it alone tonight. Please talk to a person now: the numbers under this message are free, and someone answers at this hour. If you're in danger right now, call 112. I'll still be here afterwards, for the small stuff.

Under her answer the app shows the helplines for your country, 16 countries so far, and they stay one tap away under Family. The check for danger runs on every message before the model answers, so it works even when the AI is slow. Mila doesn't diagnose and doesn't play therapist. She says plainly that it's time for a real person, and shows you one.

What she knows before you type

Your children's first names and ages. The pattern from your family check. The step you're on this week on your map. Your last conversations with her.

That's why her answer isn't ChatGPT's. ChatGPT doesn't know that bedtime is your weak spot, that your nine-year-old answers «fine» or that you're on step two. Mila does, and she starts from there instead of from zero.

She answers in English, German or Russian, whichever the app is set to.

What she does

Runs the family check: five minutes, no wrong answers, pause anytime. Each parent answers, and the kids answer their own version.

Names your family's pattern and builds the twelve-week Parenting Map.

Answers «Is it normal that…?» at any hour, in plain words, with a clear line on when to see a pediatrician or a psychologist.

Comes back after a month and asks how things are now.

What she is not

Not a doctor, not a therapist. She doesn't diagnose. If you tell her something serious, she says so calmly and shows you real people you can call, right away.

Not the product. The plan, the cards, the games, the points and the calendar are what you use every day. Mila is the guide.

How often

Free: 3 messages a day, enough for one evening's question. Plus: as many as you need, for both parents. When the AI isn't reachable, she still answers from the «Is it normal that…?» library in the app.

Your words stay yours

Mila runs on Claude, a model by Anthropic. Your own chat and a short family context (first names, ages, the pattern, the current step) go to Anthropic to generate the answer, nothing else. Anthropic doesn't use it to train models. Your partner sees the family pattern and the map, never your chats. You can delete them anytime, with one tap.

Next step: meet Mila in the free family check. Five minutes, no card.