A desktop app that hears both sides of the call, transcribes it live, and tells you what to say. The meeting never sees it.
macOS · Windows coming soon
Natt sits above your call, not inside it
It takes system audio and your microphone directly — no screen-sharing dialog, no shared tab, nothing in the Dock. The panel floats over whatever is in front of you, and answers either on its own or the moment you ask for one.
No sharing dialog to accept, no tab to share, no Dock icon. The window is hidden from screen recording and screen sharing, so it stays out of anything the other side sees.
System audio and your microphone are captured at once, so everyone who speaks lands in the transcript. No phone propped against the speakers.
Auto mode follows the conversation and replies the moment a question lands. Turn it off and the answer comes on a hotkey instead, pressed from whatever app is in front — either way you keep looking at the person asking.
One shortcut captures the screen and returns a worked solution — code, a maths problem, a multiple-choice question — using the context of the meeting.
The live transcript is translated into any of ten languages, so a call in a language you read slowly is still one you can follow. The suggested answers themselves come back in English or Russian.
Three simple steps
Download the app, sign in, and allow it the microphone and system audio. Two minutes, once.
Add your CV or bio in the app, once. Every answer after that is built from your real experience instead of a template.
Natt transcribes and translates as people speak. Let auto mode answer by itself, or press the hotkey when you decide you want one.
Your experience, not a template
Other meeting AI tools answer from the question alone, so every reply reads like a template. Aqtopus answers from your own documents — a CV, a bio, whatever you upload — so the projects, the numbers and the wording are yours.
Your CV, bio or notes — PDF or DOCX, up to ten files. Added once in Natt, and every answer after that can draw on them.
The whole document never goes into the prompt. For each question Aqtopus pulls out the passages that match it and sends only those, so the answer stays about what was asked.
The model is held to what your documents actually say — no invented employers, dates or projects. What it claims about you is something you can stand behind when someone digs in.
Recordings, on any operating system
Upload a call you already have, or record one in your browser. This half needs nothing installed — Windows and Linux included.
Automatic speaker diarization splits the transcript by voice, so a two-hour call reads like a script.
The decisions and the follow-ups, pulled out and ready to paste into a message.
A recording you keep stays in your account until you delete it. A file uploaded only to be analysed is removed from storage the moment processing finishes.
Frequently asked questions
Pay by the credit, for what you actually use. A 30-minute meeting runs about 60–70 credits.