Real-time answers from your real experience.
CueFlect listens to your interview, catches the question, and streams a concise answer built from the profile and stories you wrote — into an overlay only you see.
How an answer is produced
Four stages, while they speak- Capture System audio or your microphone, taken by the app itself. On your device
- Transcribe Whisper turns the audio into the question being asked. Device, or cloud by edition
- Answer A model answers, grounded in your saved profile and stories. Device, or cloud by plan
- Read The answer streams into a stable, always-on-top overlay. On your device
You decide what leaves
-
Audio capture, in every plan.
Local, Mixed and Cloud all capture on the device, and no plan streams your room. The thin Cloud edition sends out only detected speech segments — to be transcribed, and nothing more.
-
Session history, encrypted on disk.
Transcripts, answers and screenshots are written to your own machine, not to a CueFlect server.
-
Screenshot pixels, on Local and Mixed.
Device only. On Cloud they may be sent, as the terms of use set out.
-
Text reasoning is the part that can leave.
Local answers on the device. Mixed and Cloud send the question and your approved profile text to the CueFlect gateway or OpenAI.
Answers you can audit
Every answer names what it was built from — and what it had to leave out. Nothing is invented on your behalf: the instruction against fabricating your experience rides in every prompt, and when a prompt cannot fit, CueFlect refuses to send it rather than let the model server cut that instruction away.
Why this answer?
Illustrative. The real panel lists your own labels, never the prompt text.
On your screen, not in the share
Prevent UI Screen Capture keeps the overlay and its dialogs out of supported screen-share and recording paths — Windows capture affinity, switched by you. Try everything free before your next interview. No card needed.