What does AI see in rotli?

rotli works without any AI at all. When you do turn it on, two simple controls decide what a model can read and what it can change, and they work the same way everywhere in the app.

Two kinds of model

• On-device models run on your Mac. Nothing you ask them leaves your computer, and they work without a network.
• Connected models are tools you already use, like Claude Code or Codex. rotli labels them plainly as remote. Their own data policies apply to what you send them.

Two controls

Secure is about who can read. A note marked secure is never sent to a remote model: not its text, not its title, not even a search hit. On-device models can still read it unless you switch that off too, for one note or for the whole vault. Notes that look like they hold a secret (a password, a key) are treated the same way.

Locked is about who can edit. Any model can read a locked note, but no model can change it.

The two are independent: a note can be secure, locked, both, or neither. Mark either from a note’s right-click menu.

What secure is not

Secure is a rule about what rotli shares with AI. It is not encryption: the note is still an ordinary file in your folder. If you need the file itself protected, use your Mac’s disk encryption (FileVault) as well.

No AI at all

Leave the models off and rotli is still a complete workspace. Writing, linking, tasks, search, and views need no model and no connection.