What leaves your computer, feature by feature

Most products ask you to take their privacy claims on trust. Minda keeps a written privacy audit: its own network traffic, feature by feature, with the code behind each claim. A short version lives in the app at Settings → About → What leaves this computer.

The audit covers what Minda itself sends. The sites you open still see what any browser shows them.

Sent in the background

A handful of things happen on their own:

• Filter and warning lists are downloaded, every 24 hours for ad blocking and every 12 hours for harmful sites. No page address goes with them.
• Account refresh, about hourly while you are signed in, with your session token and the app's version, system and processor type.
• Update check: the version, release channel, system, processor and package type. No cookie, account, token or install id.
• Search suggestions: the words you type in the address bar go to your chosen search engine, without cookies, never for an address, and never from a private window. You can turn them off in Settings → General.

Sent only when you use a feature

The rest happens only when you use something:

• AI chat goes to the provider you picked, or to Minda AI, with your question, the chat so far and the pages or files you added. A model in Ollama stays on your computer.
• Sync sends your items sealed. The server still sees how many items you sync, their sizes and when they change.
• Teams, community skills, feedback and Minda Live each have a row in the audit that says what is sent.

Feedback shows you the whole report first and sends nothing until you press Send.

Controls you have

• Keep AI on this computer (Settings → AI → Privacy) makes Minda refuse every cloud AI provider before anything is sent.
• AI activity lists every model request by provider, model, host, size and page titles. The words themselves are never kept.
• Temporary chats are never saved.
• Forget browsing data clears a site, a workspace or everything.

The honest gaps

The audit also lists what Minda does not protect, and that is what makes the rest believable. A few examples:

• Hide my IP covers browsing only.
• With sync's default key, Minda's server can read what you sync. A passphrase makes it end to end.
• Saved AI chats and history are plain files, readable by anyone who can use your computer's account.
• A free proxy location is a stranger's server.

Read the audit before you rely on a claim about any one feature. The rows are short, and each one names the code behind it.