Select text anywhere and get an instant edit, hand your files to a agent, let it drive your desktop and browser, and turn a meeting into minutes — all on noit's own self-hosted models, no cloud vendor.
we was hoping to circle back on the deck before eod
I was hoping we could revisit the deck before end of day.
Everything it does
Text, files, the desktop itself, the browser, and your meetings — each behind its own permission, none of it routed through a cloud AI vendor.
Select text in any app, hit the hotkey, and a popover streams the result. Type a request instead and noit routes it to an edit, an answer, or a full agent run.
we was hoping to circle back before eod
I was hoping we could revisit this before end of day.
Sandboxed file tools plus a bulk run_shell and open_app — up to four runs at once, tracked by floating orbs.
$ pytest -q
42 passed in 3.1s
Mic and system audio merged into one 16 kHz recording, optionally transcribed, then turned into LLM-generated minutes.
A separate vision model reads the screen and clicks, types, and scrolls where the other tools can't reach.
A Chrome extension lets noit navigate tabs and act on the page — cleaning up its own tabs when it's done.
Every delete, overwrite, shell command, and consequential click pauses for your Allow or Deny.
One brain, every surface
The same self-hosted models drive each capability — and reach into the apps you already use.
Rewrote your selection and replaced it in place — one clean sentence, grammar fixed.
Chromebrowser control
Browser
tab