History and memory
Two separate things that both amount to "you can see what it did, and undo it".
Document history
MindMax keeps its own history of the notes in your folder — your own edits and the AI's alike — independent of any Git or sync you already use, and with nothing to set up.
File → Browse Version History… on an open document lists its versions, each labelled with who made it (you or MindMax AI) and what kind of change it was: an edit, an AI change, a translation, a restore. Select one to see it:
- Compare with current — what restoring would add back (underlined) and remove (struck through).
- Full text — the version as it was.
Restore this version overwrites the document with that version and records the restore as a new entry, so a restore can itself be undone. Rolling one document back does not touch anything else.
This is what makes it reasonable to let an AI write into the folder that holds your real work. If you are in the habit of duplicating a file before letting a tool near it, you can stop.
It is a safety net for changes made inside MindMax, not a backup. Keep backing up your folder the way you already do.
What the assistant remembers
As you work, the assistant distils notes about how you work — preferences you have stated, conventions you keep correcting it on, what you are working on — so you do not re-explain them every session. It learns mostly from what you do: revisions you accept or reject, outline changes you make.
All of it is plain text you can read and edit, in Settings → Memory:
- Two levels. Memories for the current knowledge base stay with it. Global ones — the language you write in, the tone you want — apply across all your knowledge bases.
- Two kinds. Tool memories shape how tasks come out. Growth memories record the topics you keep coming back to and — only after the assistant asks you — how your writing or reading is developing. They are never mixed into ordinary answers; only coaching-style work, such as a reading guide or a review of your writing, draws on them.
- Editable entries. Change an entry's text, limit a tool memory to particular tasks, or delete it. Each is a file, so you can also open it in an editor; changes take effect on the next task.
To stop it altogether, switch off Agent memory in Settings → General. Nothing new is captured and nothing is used; what is stored stays until you delete it.
Worth a look once a month if you use it heavily: a wrong conclusion left in place keeps influencing answers until you remove it.
Translation memory
Terminology settled in one translation is kept and offered in the next one in the same knowledge base and language pair. Edit it in Settings → Translation memory — see translating a document.