Skip to main content
Alice.In does not run an AI agent for you. It connects to the Claude Code you already use, on your machine and your own subscription. The app’s Connect a repo panel walks you through it and lists what is already connected. Connecting adds two plain files to your project, which you can read, edit or delete at any time:
  • .mcp.json — registers the connection to your workspace.
  • A block in CLAUDE.md — teaches your agent the protocol: orient from the project record, log what it does, pull queued work, and propose changes for your approval.
Nothing runs in the background, and nothing about your existing setup changes.

​
What changes in your sessions

It arrives oriented

Say “get the alicein context” and your agent knows the whole project state — phases, active work, recent decisions, what is waiting on you.

It leaves a record

Work normally. Your agent logs what it did and why into your project record, so nothing you build stays unexplained.

It picks up queued work

Press Queue for agent on a card, then say “check my alicein queue”. The card arrives with its full context pack.

It proposes, you decide

Changes to plans and decisions land in Needs you for you to ratify. Nothing structural changes without you.

​
Using a different tool

Any card offers its context as markdown you can copy and paste into Cursor, ChatGPT, or whatever you prefer. You lose the automatic logging, not the context.

​
Your work shows on the map

Once connected, the commits your agent records are matched to your modules by the files they changed, so recently worked parts of your system read as active on the factory map. You can connect more than one repository — Connect a repo in the app shows what is already connected.
We never see or spend your Claude account, and we never ask for its credentials. The connection carries your project’s record, nothing else.