The hosted chat that used to run on this page is switched off. Everything it did now runs on your own machine instead: still no account, still no login, and no token allowance to run out of.
Windows, macOS and Linux. Free forever, and the source is public.
What changed
The anonymous chat on this page ran on our servers against a shared token pool. That is the part that is gone. SurfSense itself is not going anywhere: it is a desktop app now, and the app is the free version.
That swap is why there is no allowance any more. We are not paying for your inference, so there is nothing to meter: you bring a key and your provider bills you at cost, or you run a model on your own hardware and pay nobody. Either way the cap that used to cut you off does not exist.
In the app
Qwen3 is in the app's own model catalog. Pick it, let the app download it, and it runs on your hardware from then on, with no key and no network connection.
If you want a frontier model for the hard questions, add an OpenAI-compatible connection with your own key and switch between the two per chat.
Add PDFs, Word files, spreadsheets and pages, ask across all of them with citations, and turn the answers into a deck, a report or a briefing podcast.
The hosted chat was free until the shared pool ran out, which is the part nobody liked. The app has no pool to share: it runs on your machine, so the only limits are your own hardware and whatever your provider charges you.
There is no seat to provision, no tenant to configure and no admin console, because there is no server. The only identity involved is your operating system user.
Not on our servers: the hosted free chat that used to run on this page is closed. The SurfSense desktop app has no account and no login at all, so nothing there asks you to sign in. Qwen3 is in the app's local model catalog, so it downloads and runs on your own hardware.
Yes. The app and every update are free forever on Windows, macOS and Linux, and a 30-day licence for the scraper plugins comes with it. There is no account to create and no trial clock on the app itself.
They were a shared allowance on our hosted service, and that service is switched off. The app has no allowance because we are not paying for the inference any more: you either point it at your own provider key, which your provider bills you for at cost, or run a model on your own hardware for nothing.
No. Qwen3 is one of the models the app downloads into its own local catalog, so it runs on your machine with no key and no network connection. You can still add a key for a hosted model later if you want one.
Nowhere. Sources, chats and everything generated from them live in one local workspace, a folder in your user directory with a database file in it. We hold no copy and keep no log, and with no account there is nothing tying you to one. Anonymous chat was never stored in a database either, so there is nothing of yours left on the old service.
One installer, no account, and your documents never leave the machine you put them on.
