What we collect — all of it.

Most privacy pages give you vague categories. This one gives you the actual list: every field Cortex ever sends, what it's for, and how to turn it off.

BETA

During the beta, telemetry is on by default.

A beta exists to find crashes and regressions, and the counters below are how we see them. So while Cortex is in beta, this anonymous telemetry is on by default — and unlike most browsers, the off switch is total: flip one toggle in Settings → Privacy and nothing is sent at all. Not reduced. Nothing.

When Cortex leaves beta, telemetry will be off by default.

The full telemetry list

Sent at most once per day, as a single small batch of counters and averages. This is the complete list — there are no other fields.

Identity: a random ID, and nothing else

A random install ID (a UUID generated on your machine — not derived from your hardware, not linked to any account or email), plus the app version and your operating system. That's the entire identity picture: we can tell two installs apart, and nothing more.

Stability counters

Renderer crashes, main-process errors, failed tab restores. The reason the beta collects anything at all — without these, a crash on your machine is invisible to us until you happen to report it.

Memory Saver counters

How many tabs were suspended and how much memory that freed. This verifies the headline feature actually delivers on real machines, not just ours.

Performance averages

Average CPU and memory use, average and 95th-percentile page-load time, startup time, and a tab-count snapshot. Averages only — never which pages, never which tabs.

Feature adoption

How many contexts exist, switch counts for the four built-in contexts, and yes/no flags for whether sync, AI, Your Rhythm and Dark Reader are enabled — plus how many contexts have ad-block on, and a count of trackers blocked. Tells us which features earn their keep.

What the records don't contain

Telemetry rows hold no IP address, no timestamps of individual actions, no per-site anything. Custom contexts you create aren't named in switch counts — only the four built-in ones are counted. One counter — trackers blocked — keeps counting locally even with telemetry off, purely to show you your own number on the new-tab page. It never leaves your machine either way.

What we never collect

Not during beta. Not after. Not "anonymized". Just not collected.

URLs & browsing history Page contents Searches Keystrokes Geolocation Ad IDs or fingerprinting Any link between telemetry and your account

The rest of the data story

Feedback

Only what you type, only when you press send. Message, category, app version and platform — nothing attached, nothing linked to telemetry or your account.

Sync — opt-in, end-to-end encrypted

If you create an account, bookmarks, contexts and settings sync as an encrypted blob our server cannot read — the key never leaves your machines. No account, and nothing syncs at all.

Your Rhythm — never leaves your machine

The opt-in work/relax insights feature computes everything on-device. It is not part of telemetry and is never uploaded, on any setting.

AI — your keys, your provider

The assistant talks directly to the provider you configure (or a local Ollama, in which case nothing leaves your machine). Your prompts never pass through our servers.

Everything above lands on our own self-hosted server — no analytics companies, no third-party services touch it. And none of it is ever sold or monetized. That's not a policy that can quietly change with an update; it's the reason Cortex exists. Read the philosophy →

Comfortable with that trade?

Join the waitlist →