Data and privacy
Here is exactly what Cloud OS does with what you tell it. No more, no less.
What you own
Your memory is saved as files in a private repository in your own GitHub account. Cloud OS writes your memory to that repository, so the memory we keep is never the only copy.
What Cloud OS keeps
We keep one search index per person in our database and a copy of your repository on our server. Both hold the text of your pages as plain text, because search has to read it. Our app database also holds your answers while setting up and short summaries of what changed in your memory. Deleting your account in Settings removes all of it from our live storage. Encrypted backups of our app database are kept for about a week and then pruned. The search index has no backup, because it rebuilds from your GitHub. Your answers as you typed them exist only here, so they are gone for good; the pages of your memory stay in your GitHub.
What goes to an AI provider
Setting up, search and summaries run through OpenRouter, so relevant text passes through an AI provider on its way to an answer. While you set up, Cloud OS's requests go only to providers that keep nothing they're sent. Our OpenRouter account is set to use providers that don't train on your data.
How it's protected
Backups are encrypted. The GitHub access we use to sign you in is stored encrypted in our database. The access the server uses to read and write your repository sits in a private file on the server, not encrypted. You can only sign in with GitHub, and only when you're invited.
What we don't claim
Cloud OS isn't end-to-end encrypted: a server that searches and summarises your memory has to be able to read it. The stronger guarantee is running the same open-source engine yourself.