It's 4:00 AM
project4am is named after the hour we want you to stop. A self-hosted assistant that handles your customers in your voice, so you can finally close the laptop.
04:03 · the problem
You're the CEO, the support agent, the copywriter, and the coach who trains every new hire. You answer the same question for the hundredth time. You tried a generic AI to help, and turned it back off, because it sounded cold, corporate, nothing like you. So you went back to doing it all yourself. At 4am.
04:17 · the idea
Not a chatbot pretending to be a person. A second you, your voice, your judgment, your warmth, that never sleeps, never gets tired of the same question, and always sounds like you.
How it works
Drop in your site, docs, and FAQs. It chunks, embeds, and indexes everything, and cites the source on every answer, so it never guesses.
Paste a handful of replies you would actually send. It learns your phrasing, your warmth, your "anything else?", until it reads like you wrote it.
It handles the routine, books appointments, captures leads, and hands the hard ones to you, neatly queued for the morning.
The difference
We built it around the one thing generic AI gets wrong: you.
From real examples of how you actually reply, even the playful, human "throwing it in your tote" lines that make it sound like you.
Every answer is grounded in your own knowledge and cited. If it doesn't know, it says so and checks with you. It won't guess.
An excited new customer, an anxious first-timer, a frustrated complaint. It matches the energy instead of replying like a form letter.
Thumbs-up a reply and it teaches your team. In a few weeks it's calibrated to your standards, not generic defaults.
Generic AI vs. your second self
Everywhere your customers are
Messenger, Instagram, WhatsApp, and a web widget for your site: one assistant across all of them, in five languages. It books appointments and captures leads while you sleep.
From people who stopped at 4am
For the first time in two years I closed the laptop at 6pm. It answered 38 messages overnight and not one of them sounded like a robot wrote it.
We're an apothecary, so tone matters and bad advice is dangerous. It defers to a human exactly when it should, and never invents a fact.
Bookings come in across WhatsApp, Instagram, the widget, in three languages, and land in one place, already drafted in my voice. It paid for itself in a week.
This isn't hypothetical
project4am didn't come from a boardroom. It came from the daily reality of small businesses drowning in messages: owners who answer every customer personally, who tried a generic bot and switched it off because it didn't sound like them. It's already handling those evenings back.
07:00 · not 4am
You wake at 7, not 4. Over coffee you open the dashboard: 47 conversations came in overnight. It handled 41 on its own, friendly back-and-forth, a few bookings captured. Six are waiting for you, already drafted in your voice. You review them in fifteen minutes. Your team didn't ping you once. You close the laptop at 6pm.
Built for you, not Silicon Valley
Self-hosted
You run it on a server you control. That's the whole cost: the price of a few coffees, scaling with your volume. No per-seat fees. No surprise invoice.
Most AI assistants cost $100–500 a month, a deal-breaker when your monthly profit is $200. We think your second self should cost less than your coffee.
Get started →Questions, at any hour
It's 4:00 AM
Spin up your second self in an afternoon. Close the laptop tonight.