Your screen, composed for the moment.

Embreon is a personal AI assistant. In the morning, it prepares your briefing. While you work, it steps back. And before acting for you, it asks for your OK.

One email at launch, nothing else. How your address is used

Work in progress: the interface is designed, the product is being built episode by episode.

A day with Embreon

  1. In the morning, it composes your briefing.

    Calendar, emails to answer, pending pull requests, weather: Embreon builds the screen around what matters today, and every card says why it’s there.

    Morning screen: Embreon’s core speaks, with four cards for calendar, email, code and weather, and the overnight log.
  2. While you work, it steps back.

    It keeps your context in view, holds notifications until your focus block ends and fixes the failing CI in the background, without pushing anything without you.

    Focus screen: the “Voice pipeline” block, the failing test, the coding agent at work and seven held notifications.
  3. Before acting for you, it asks.

    An email going outside? It shows you the draft, explains why it’s asking, then gives you 30 seconds to undo after you approve.

    Alert screen: an approval request to send a quote reply, and the sprint review starting in five minutes.

You stay in control

An assistant that acts for you has to be easy to control. Four rules, from the first version.

Everything is logged.
Every action goes into a log, with the time and the reason.
You set the autonomy.
Careful, Balanced or Autonomous: you choose what it does on its own.
Nothing leaves without your OK.
Emails, messages, payments: in Careful mode, it always asks.
What can be undone, can be undone.
Reversible actions keep their Undo button, right in the log.

Built in public

I’m Alistair, a developer. I’m building Embreon episode by episode, and the code will be open source.

One email at launch, nothing else. How your address is used