Project context, so you can switch AI without starting over

While it works on a project, the AI writes down what it knows about it: the project context. So if you switch tools tomorrow (from Claude to Cursor, to ChatGPT or to Bitácora's built-in AI), the new one doesn't start over: it reads the context when it connects and keeps working the way the previous one did.

What it keeps

What the screens don't show and is needed to carry on the same way:

• The visual system and the reasons behind it: the colors, the type, the spacing and why they were chosen.
• The interactions: what happens when you tap something, what animates and which states there are.
• The data: what repeats across the project and how it's presented.
• The decisions: what was agreed and shouldn't be undone.

It's split by topic, one file each (for example, visual style, interactions or data). It only holds design knowledge: nothing from your account.

You don't have to do anything

Bitácora makes sure the AI uses it:

1. When it opens a batch, the connector asks it to read the context before touching anything, and it works in line with it.
2. While it works, if it changes something the context describes or makes a decision the next session should know about, it updates the context in the same batch.
3. When it closes the batch, if it changed the project without reviewing the context, Bitácora asks it to review it before finishing.

It works the same with any connected tool and with the built-in AI.

When it appears

A new project has no context. It starts when there's something to tell: the first time the AI settles a visual system, an interaction pattern or a piece of data that repeats, it writes it down.

See it or correct it

For now the context isn't shown in Bitácora. If you want to know what it says, ask your AI in the chat ("show me the project context"). If something isn't right, tell it, or dictate it in a note, and the AI corrects it.

Each project can have up to 30 context files, up to 64 KB each.