Trippi Cue questions, answered

A Chrome extension for interviews that happen in a browser call. It hears the question addressed to you, shows a short suggested answer while the other person is still waiting for it, and translates the conversation underneath.

Whoever is in the chair being asked things: candidates in a technical, behavioral or case round, people interviewing in a language they know less well, and anyone who freezes for three seconds before a good answer arrives.

Those work before the call — mock rounds, question banks, feedback afterwards. Cue does nothing before and nothing after; it works in the twenty minutes you are actually being interviewed.

No. Trippi puts translated captions over a whole meeting so you can follow a language you do not speak. Cue answers the question asked of you and translates the call underneath. Separate extensions, separate plans.

One line helps: “About this meeting” takes a sentence about the role and the round, and every answer is written against it. Nothing to upload, nobody to invite.

A cheap local check drops the small talk, then a model looks at the recent lines, the participants and your name and decides whether the closing turn was addressed to you. On a panel, questions put to somebody else never become cards.

Three reasons: the model judged the question was not yours, its confidence was below the threshold, or the same question already had a card in the last minute. A quiet window is usually the product working, not failing.

Yes. There is a field at the bottom of the window: your question goes off with the last minutes of the call as context and skips every automatic filter.

Yes. It is written by a language model from what was said, and it knows nothing about you beyond the call and your one line of context. Read it before you say it — it is a prompt for you, not a script.

A few seconds after the question is finished, which is the point: an answer that arrives after the topic has moved on is not an answer.

Three by default, five at most. The window is meant to be glanced at mid-sentence, not read like a feed.

Yes, and it does by default. Every recognised line also arrives in your language, from the first minute — 68 to choose from.

That is the normal setup: the question stays as it was asked, the answer is written in the language you picked. You can be interviewed in German and think in Polish.

Yes — set both language buttons in the window header to the same language, and the lines stay exactly as they were spoken.

No. The picker works mid-call and the copilot keeps running while you change it.

Google Meet, Microsoft Teams on the web and Zoom on the web, in Chrome and Chromium browsers on a computer.

A browser extension can only reach browser tabs. The desktop applications are separate programs — no extension can see inside them, ours included. Ask for a browser link, or use “Join from your browser”.

No. Mobile browsers do not run extensions like this one, and a call in a phone app is not a browser tab at all.

Yes, Cue recognises the Defender for Cloud Apps addresses. There it works on the click-only permission: start it by clicking the icon once the call is open.

No. Nothing appears in the participant list, nothing announces itself, and the other side sees exactly what it would see without Cue.

Not from the call itself. But the window is drawn in your browser page, so it is visible to everyone if you share that tab — and we promise no secrecy beyond that plain description.

That is between you and whoever is running them. Many assessments say plainly that outside help is not permitted, and Cue is outside help. We do not build anything aimed at proctoring software, and we say so on the responsible use page.

No. We have no relationship with any of them and there is no integration that could.

The audio of the meeting tab, and only after your click. It never touches your microphone, your camera or any other tab.

Straight from your browser to our speech-to-text provider over an encrypted connection. It does not pass through our servers, and it is not recorded anywhere.

A short rolling window of the recognised lines, the speaker labels, your name and your one line of context — enough for the model to tell whose question it is and to answer it.

The recognised text of the call and the answers are kept for 30 days, so a poor answer can be told from a poor question, and are then deleted automatically. They are never used to train a model and never sold, and deleting your account deletes them.

No, and it cannot be: the audio is never written to disk on your machine or ours, and no file or summary of your call is produced at any point.

No. Cue hears speech. A shared diagram, an editor, a slide — none of it reaches the extension, which matters most in coding and system design rounds.

Not today. There is nowhere to upload a file, and nothing would read it — answers come from what was said and from the line you typed in.

No. Nothing is produced after the call: no notes, no score, no advice on how you did.

No. It works in a real call, with a real person asking real questions.

Yes, and it does not expire. The monthly allowance renews on its own, and the window shows what is left of it.

Hours of call that Cue listens to. Answers are not charged for — a round where it stays quiet costs the same as a busy one.

Cards stop and the window says so. The call is untouched. You can move up a plan mid-month or wait for the first of the month.

Cancel any time from your account page; refunds within 14 days of the payment, no argument required.

Which platform the call was on, what you expected, what happened, and whether the window opened at all. If you write from the extension, your version and platform arrive with the message.

Usually within one working day, and a person answers — there is no ticket queue to get lost in.

Please do. What people ask for is how the list gets longer.