UnderstudioCreate accountYour account

The answer is ready before you finish reading.

Understudio is a desktop copilot for technical interviews. It listens to the call, reads your screen, and puts a worked solution one keystroke away.

Create an accountSee pricing

macOS app coming soon
macOS 13+ · free to install · runs on tokens, from $9 a pack

Illustrative session
  • INTERVIEWERLet's start with something small before the system design part.
  • YOUSounds good.
  • INTERVIEWERWalk me through how you'd find two numbers that add up to a target.
  • YOUSure — I'd use a hash map so it's a single pass instead of nested loops.
  • INTERVIEWERAnd what does that cost you in memory?

The conversation becomes the question.

Press ✦R and everything said since the session started goes with it, stopping only at a pause longer than a minute. Every answer it already gave you, the screenshots you queued and the parts of your notes that match go in the same request. No copy-paste, no retyping.

Microphone and system audio feed the conversation; with earlier answers, pending screenshots and matching note sections it forms the context. Pressing the Hyper key and R turns the context into an answer, and every answer goes back into the context for the next solve. every answer goes back inMicrophoneYOU · your channelSystem audioINTERVIEWER · the callScreenshots✦H · queuedYour notes.md · matching sectionsContextThe conversation so farEarlier answers, this sessionPending screenshotsMatching note sections✦RAnswerapproachOne pass, hash mapcodedef two_sum(nums, t):seen = {}complexityO(n) time · O(n) space

  1. It hears and sees

    • Microphone YOU · your channel
    • System audio INTERVIEWER · the call
    • Screenshots ✦H · queued until you solve
    • Your notes .md · matching sections

  2. Context

    • The conversation so far
    • Earlier answers, this session
    • Pending screenshots
    • Matching note sections
  3. ✦R

  4. Answer

    • Approach, code and complexity
    • Goes back into the context for the next solve

Three keys, start to finish.

  1. Hyper key and S: Start the session

    Name the round and start. Both sides of the call are transcribed, every line marked You or Interviewer.

  2. Hyper key and H: Capture what matters

    Grab the editor, the diagram, the whiteboard. Shots wait in the feed as pending context until you solve.

  3. Hyper key and R: Solve

    A worked answer lands in the panel — approach, code, complexity — and stays there to scroll back to.

Two keys instead of four.

The defaults sit on ⌃⌥⌘ so ⌘S, ⌘H and ⌘R keep working in every other app, which makes each hotkey four keys. Turn on the Hyper key, and holding Caps Lock (or a key you pick) stands in for all three. Solve becomes ✦R.

On macOS, and off until you switch it on. It needs Accessibility, which macOS asks for when you do, or it can use the Hyper key Raycast or Hyperkey already gives you. A quick tap of Caps Lock still turns caps on and off. Turning it on and off

18 hotkeys in all, and every one but Settings can be rebound. The full list.

Out of the way until you need it.

Floats over the call
Its own window above Zoom, Meet, Teams or a browser tab, on every desktop and in full screen. It captures system audio instead of hooking into the call, so it doesn't care which app the call is in.
Fades back and lets clicks through
Dial the panel down and turn on click-through, so it never steals a click from the editor underneath.
opacity 30–100% · ✦T
Both sides, labelled
Your microphone and the call's audio are transcribed as separate channels, so every line is marked You or Interviewer by where it came from.
Screenshots, batched
Capture as many frames as the question needs. Solve one on its own, or send every pending shot in as a single problem.
✦H
Follow-ups land on the right answer
Every answer from the session goes along with the next solve, so “make that iterative” or “what's the complexity of that” knows what “that” is.
Your last 50 rounds, on your Mac
Transcripts, screenshots and answers are saved on your own disk as you go, not on our servers. The most recent stay; screenshots keep as thumbnails.
50 sessions

Off their screen, as far as macOS allows.

From the first launch, the window asks macOS to keep it out of screen shares, recordings and screenshots. What you share shows whatever is behind the panel.✦V turns it off and back on.

Where it stops

macOS treats this as best effort. On macOS 15.4 and later some capture tools can still see the window, and the app can't tell when that happens: the toolbar's Hidden label shows the setting, not proof. Try a test call with the tool you'll use before it matters.

Your screen
The shared screen

Brief it before the call starts.

Drop in a Markdown file: the take-home you already solved, the role brief, your own notes on the system you built. It's indexed and retrieved alongside the live transcript, so answers land in your context instead of generic ones.

  • Attach as many files as a round needs
  • Every answer shows which file and section it drew from
  • Kept on your Mac; only the matching parts go out with a solve

Pick a monthly allowance.

Run out mid-season? Top up with tokens instead of upgrading.

Pro

$25/ month

Billed monthly · cancel anytime

For a focused job search — a handful of rounds a week.

Start with Pro
  • 500 tokens per month
  • Last 50 sessions, stored on your Mac
  • Live transcription with speaker labels
  • Markdown context files per session
  • Screenshot & batch solve
  • Overlay mode, all hotkeys
  • Token top-ups

Max

$65/ month

Billed monthly · cancel anytime

For back-to-back onsites and the hardest technical loops.

Start with Max
  • 1,500 tokens per month
  • Everything in Pro
  • Frontier models & longer context
  • Custom prompts per session type
  • Auto-solve on new questions
  • Priority support

A standard solve is about 1 token, a frontier-model solve 2–4, and an hour of live transcription about 12.

Token packs never expire.

Pack tokens are spent only after your plan's monthly tokens run out, so a busy week never means an upgrade. You can also buy them with no plan at all.

  • 250tokens$9$0.036 / token
  • 1,000tokens$33$0.033 / token
  • 2,500tokens$80$0.032 / token

Good to know.

Still unsure? Talk to us — we answer within a day.

Does it work with any calling app?
Yes. It runs as its own window alongside whatever you're using — Zoom, Meet, Teams or a browser tab — and captures system audio rather than hooking into the call.
What does a token buy?
A solve is one press of ✦R. A standard solve is about 1 token, a frontier-model solve 2–4, and an hour of live transcription about 12.
What happens when I hit my monthly limit?
Pack tokens are spent automatically once your plan's run out. If you have none, you'll be prompted to top up — until you do, there are no solves and no live transcription. Plan tokens reset every month.
Where is my data kept?
On your Mac. Transcripts, screenshot thumbnails and answers are saved there, not on our servers. While a session runs, audio goes to the transcription service; when you solve, that context goes to the AI model. Neither keeps it or trains on it. The privacy policy has the details.
What does signing in with Google or GitHub share?
Your name, email address and profile picture, plus your username on GitHub — that is the whole list. They identify your account and let us send receipts. GitHub's email permission covers every address on the account. We check them to pick your primary one and to find an Understudio account you already have; only the primary is kept. Understudio never asks for access to your mail, files, calendar or repositories, and holds no permission that would let it.
Can I cancel or switch plans?
Any time, from Billing & usage. A cancelled plan runs to the end of the period you paid for and is not charged again. Upgrades take effect at once; downgrades wait for the next renewal. Pack tokens stay yours either way.

Your next round starts soon.

The macOS app is not open for download yet. Create your account now, and the app signs in with it the day it lands.

Create an accountLog in