Skip to main content
Steps 1–5 happen entirely in your browser — you get a working product even if you never connect an agent. The last step connects the AI agent you already use, and that is where it gets good.
1

Open your workspace

We send you a link and a token. Keep the token like a password — it is what identifies your workspace.Your workspace holds your project’s live map and its memory: everything decided, everything done, everything pending.
2

Tell it what you are building

Describe your project in your own words, and give us your GitHub repository URL. Public repositories work as they are; a private one needs an access token and we will walk you through it.You are not filling in a project management tool. You are describing the thing, once.
3

Watch your workspace get built

In about a minute, two things exist that you did not have to draw:
  • A starter plan — goals, phases and enablers, inferred from what you described. This is a first draft with your project’s actual shape in it, not an empty template.
  • A factory map — your services, libraries and external dependencies, read out of your code and laid out as a place you can walk through.
Expect the starter plan to be roughly right and specifically wrong. That is the point — correcting a draft takes minutes, and starting from a blank page takes a week.
4

Correct it, and make it yours

Open the project map and shape the draft into your real plan: drag enablers onto the phase they belong to, rename what reads badly, delete what is not real, add what is missing.Already have a plan in Linear, Trello or a document? Do not retype it — see Bringing in existing work.
5

Look around

Four surfaces, nothing installed:
  • The project map — your plan: phases, enablers and cards.
  • The factory map — your system: modules and how they connect.
  • Alice, bottom right — ask “where are we?”, “what needs me?”, “why did we do X?”
  • Needs you — the one queue where things wait for your judgment.
Press ? at any point for the keyboard shortcuts.
6

Connect your agent

Alice.In plugs into the Claude Code you already use, on your own machine and your own account. Connect a repo in the app walks you through it.Once connected, say “get the alicein context” in any session and your agent knows the whole project. Hand it work with the agent queue.

​
Your first hour, if you want a path

  1. Correct the draft. Ten minutes on the map: right phases, right names, nothing fictional left in it.
  2. Write down one decision you have already made, and why, on the card it belongs to. That is the record starting.
  3. Connect your agent, say “get the alicein context”, and ask it what it thinks you should do next. The answer tells you how good your plan is.
  4. Queue one card and let it do the work. Watch it come back for review.

​
What to expect

  • Your agent runs on your machine and your own Claude subscription. We never see it or spend it.
  • Your workspace is isolated per account. We export or delete everything on request.