Rosa reads your mail and calendar, sorts what you need, writes the replies and reminds you of what you promised. You press send.
You don't type prompts. Rosa reads your mail, calendar and Slack, tracks what was agreed, and sends you a message when something needs you: a deadline, a reminder, a departure time. The rest of the day she's quiet.
Sorted within fifteen minutes, remembered as long as it matters. At 07:00 AM you get one message with the three things you need today, instead of an inbox of forty.
"Leave in twelve minutes, traffic on the A2." "Van Dijk hasn't responded in eight days." "You promised Marc a presentation by Friday." She sends this on her own, when it matters.
She writes the reply and puts it in your drafts; you press send or adjust a line first. There is no setting that lets her send on her own.
Mail, calendar and memory live on your machine. Before Rosa asks the AI model anything, every name, number and amount is replaced. For an ISO or security audit the answer to "where is this data" is one word: here. Further down you'll see it happen.
Thirty-four messages sorted into today, this week, for info and ignore. Two flagged as things you need before noon. You haven't opened your laptop yet.
What matters, what moved, who's waiting for you, and the weather where you're driving. Delivered via iMessage before you open anything.
Your 09:30 AM appointment is on-site. She checked traffic, factored in parking time and told you when to walk out. Not when the meeting starts.
You're in a meeting. Nothing came in that's more important than the person across from you, so she has nothing to say. This is what most of the day looks like.
She follows the procurement portals you pointed her to, scores every publication against what you've won before, and blocks time in your calendar for the deadline.
You're walking to your car and something clicks. You tell her, poorly phrased, in twenty seconds. She writes it up properly, files it under the right company, finds the conversation it belongs to and brings it back when you have time.
You said it in a Slack conversation and moved on. She noted it as an open loop with a name and a date, and drafted the reply based on the real conversation and the amount of €40k from March.
Every quote and every "I'll think about it" becomes a phase in a pipeline she maintains from your conversations. If one goes quiet longer than normal for that client, she says so.
Three things didn't happen today. She tells you which, why they matter, and puts them where you'll see them at 07:00 AM instead of letting them slip.
Quiet hours are quiet hours. Unless something truly happened that you marked as urgent, she has nothing to say until morning.
Multiple businesses, six mailboxes, a calendar filled by others: we needed an assistant that knew our context. But every AI assistant we looked at wanted all our mail in its cloud. For a company that has to pass ISO and security audits that's a no. So Rosa runs on your own Mac, all data stays with you, and the model only sees anonymised text. She has read every mail we received for a year, and none of it has left the building.
We didn't want the mail of two companies on the server of a startup. So Rosa runs on your machine, or in a container that's yours alone. And before anything goes to the model to reason about, every name, number and account is replaced. See for yourself.
Tell us briefly what your day looks like. Our team reads every application personally and gets back to you within two business days, even if the answer is no. Every founding spot is installed and running by 1 november at the latest.