Setup to first answer in minutes

From your resume to a live answer, in six steps.

You do the setup once on the web. Then the Windows client carries it into the call: it listens to both sides, works out when a question lands, and hands you an answer built from your own story — on an overlay the interviewer can’t see.

The flow

One clean path, start to finish.

Steps one and two are a one-time setup. Three through six repeat for every interview you take.

01/06
You

Create your profile

In the web console you build your candidate profile — background and resume, prepared STAR stories, the job description, and what to emphasize or avoid. This is the ground truth every answer is built from.

02/06
Setup

Pair the Windows client

Download the desktop app and pair it to your account with a code. The web app stays your control plane; the client is the only thing that ever touches live call audio.

03/06
You

Join your interview

Start your Zoom, Teams, Meet, or Webex call as normal. Launch the overlay and run the one-tap self-test to confirm you are invisible on the shared screen.

04/06
Them + You

It hears both tracks

System-audio loopback captures the interviewer ("Them") and your microphone captures you ("You") on separate tracks. Live transcription runs in real time — labels are always right.

05/06
Copilot

You get points + an answer

The moment a turn ends, the copilot returns the restated question, 2–4 talking points, a first-person answer, and IMPORTANT reminders — in about a second, grounded in your profile.

06/06
Them — screen share

Stay invisible on the share

The overlay is excluded from capture, so when you share your screen the copilot simply isn’t in the frame. You glance at it; the interviewer sees only your desktop.

Why desktop-only

One live client, one control plane.

A browser physically can’t do the two things that matter most here — capture native call audio, and stay off the screen you share. So the Windows app is the only live client, and this web app is the control plane behind it.

What a browser can’t do
  • Reach the call’s system-audio loopback — a tab hears only your mic or a tab you pick, never the interviewer.
  • Hide from screen capture — a browser tab is always in the frame you share.
  • Sit always-on-top with global hotkeys over a native meeting app.
What the desktop client does
  • Taps native loopback ("Them") and your mic ("You") as two separate tracks.
  • Sets WDA_EXCLUDEFROMCAPTURE, so it is genuinely absent from the share.
  • Floats always-on-top, driven entirely from the keyboard, and self-tests exclusion.

The split is the whole design. The web app holds your account, candidate profile, wallet, sessions, and paired devices — everything you set up and review. The desktop client holds the one thing that has to run locally: the live call. Neither tries to be the other.

The loop

Transcription → turn → assistant.

The client captures the room and draws the overlay; the platform transcribes and runs the assistant — the whole loop closes in about a second. Two audio tracks in, one structured answer out, then it steps aside the instant you start talking.

copilot.loop()
  1. 01captureloopback(Them) + mic(You) → two tracks
  2. 02transcribereal-time speech-to-text
  3. 03turnend-of-turn detector fires when a thought lands
  4. 04assembleprofile + rolling transcript → prompt
  5. 05assistantQ · 2–4 points · answer · IMPORTANT
  6. 06renderoverlay — excluded from capture
# cost: ~1 credit / minute · hard-stop at zero · transcripts off by default
Two things to set up

Get the client. Build your profile.

Download the Windows app and create your account — the profile you write on the web is what makes the answers sound like you. Your first credits are on us.