> ## Documentation Index
> Fetch the complete documentation index at: https://docs.alicein.xyz/llms.txt
> Use this file to discover all available pages before exploring further.

# Memory

> What the system remembers, and why answers come with citations.

The project record is kept by the system, not by your discipline. You do not maintain it; you use it.

## What it keeps

* **Why** — the decisions, with their reasoning, in the order you made them.
* **What happened** — an append-only log of work, commits, deliveries and questions.
* **What was produced** — artifacts attached to the part of the project they belong to.
* **What was learned** — a lesson one agent learns is served to every agent, and has to cite the event that taught it.

Nothing is overwritten. When something is superseded, the old entry stays and the new one supersedes it — so the history of your thinking survives, not just its current state.

## Answers come with receipts

Ask Alice why something is the way it is and she shows you what she read. Every card carries a **History** section: the events and decisions that name it, in one place.

This is deliberate. An AI that recalls is guessing; one that cites is answerable. If the record does not know, you get told it does not know.

## Nothing goes stale quietly

Every surface knows whether what it is showing is derived live, rebuilt automatically, or written prose that is now ageing — and it says so, with an "as of" date.

Confidently wrong is worse than empty. A blank panel makes you go and look. A stale one that sounds certain does not.

## What you can do

* **Tell it to remember something.** Say it in chat and it goes into the record.
* **Read anything.** There is no hidden state — every entry is inspectable.
* **Export or delete everything.** Ask us and it is done. Your record is a git repository, backed up nightly.

## What it never keeps

Credentials. Not your Claude account, not your repository tokens, not anything you paste that looks like a secret.