macOSA voice mock interviewer that sees your screen

LeetCode checks your code. Moxie interviews you.

Open any problem: a LeetCode tab, a design doc, a whiteboard. Press Start and talk it through. Moxie listens, looks at your screen when it needs to, stays quiet while you work, and writes you honest feedback afterwards.

3 free interviews · $25 once after that · macOS 15.2+ · uses your OpenAI key

Works with whatever is on your screen

  • A LeetCode tab
  • Your own editor
  • A whiteboard in the browser
  • A SQL editor
  • A .md file
  • A pasted prompt

Why Moxie

Solving problems in silence is not interview prep.

Interviewers grade what they hear and see: how you clarify, plan, think aloud, and recover. That part only improves with a person on the other end. Now there is one, any time you want.

Thinking aloud is a skill

Moxie is voice-first. It asks for your approach before you type, and notices when you go quiet for two minutes.

Real pressure, on your schedule

A voice on the line, a clock running, follow-ups you did not plan for. At 11 pm on a Tuesday, with nobody to book.

Feedback with receipts

Every score cites what you said, when you said it, and the frame of your screen at that moment.

How it works

From open tab to feedback in three steps.

  1. 01

    Point it at a problem

    Pick a window with the macOS share picker. Moxie reads the problem off your screen. Or paste it, or drop a .md file.

  2. 02

    Talk it through

    A voice interviewer asks, waits, and follows up. It looks at your screen the way someone on a screen-share would.

  3. 03

    Read your feedback

    Scores per category with evidence, the one thing to fix first, and the moments that mattered.

01 · SETUP

It reads the problem so you don’t retype it.

Moxie reads the problem off your screen and writes the interviewer a private rubric: expected approaches, common mistakes, a ladder of hints. You check the Preview Card, then begin.

  • Coding, system design, behavioral, or custom
  • Describe your interviewer and the role you’re going for
  • A timer from the default up to 110 minutes, or off

02 · THE INTERVIEW

An interviewer, not a copilot.

Moxie never solves the problem, never names what you haven’t named, and never grades you mid-interview. It stays quiet while you work. If you’re stuck, hints come one rung at a time, starting with a question.

  • Listeningwhile you talk
  • Thinkingabout what you just said
  • Lookingat your screen, or on your hotkey
  • Speakingwith live captions you can toggle

03 · FEEDBACK

Feedback that shows its work.

Graded honestly: 3 is average, and 1s and 2s exist. Each score comes with evidence from the transcript, plus the moments that mattered, each with the frame of your screen at the time.

Built for the loop you’re in

Every round of the onsite, on one Mac.

Four Kinds of interview

Coding, system design, behavioral, or custom. Each has its own phases and grading categories. Behavioral stories get probed for context, action, result, and learning.

Any interviewer you need

Write the Persona in plain words: “Staff engineer at a payments company, hiring senior backend.” It sets the tone and the bar.

Hints that don’t give it away

The hint ladder runs from a Socratic question to near-solution, one rung at a time. Your feedback counts the rungs you used.

Nudges, not interruptions

During a long silence, Moxie decides whether to say one sentence or nothing. Usually it’s nothing.

Interview history

Every interview is kept with its transcript, feedback, and the frames Moxie looked at.

A native Mac app

Lives in the menu bar. The Panel floats over your editor and stays out of the way. One hotkey to make Moxie look.

⌃⌥LLook now

Local first

Your key. Your Mac. Your transcripts.

Moxie talks to one server: api.openai.com, with your own API key kept in the macOS Keychain. Screen frames are captured locally and sent only when Moxie looks. Every interview is saved as a folder on your disk, and nowhere else.

No Moxie account. No Moxie servers.

What it costs to run

About $2.55 for a 45-minute interview.

Paid straight to OpenAI at their rates. No markup from us.

OpenAI usage for one 45-minute interview
Voice, $0.05 per minute$2.25
Prep and feedback≈ $0.20
Interviewer judgment and looks< $0.10
Total≈ $2.55

Measured in testing with the default models. The Panel shows the running cost live.

Pricing

One price. No subscription.

Try three full interviews free. If it helps, buy it once and keep every update.

Moxie for Mac3 free interviews
$25one time
  • Unlimited interviews, all four Kinds
  • Feedback, transcripts, and interview history
  • Every future update included
  • Your past interviews stay readable, even before you buy
Download free trial

macOS 15.2 or later · needs an OpenAI API key

Questions

Can I use Moxie during a real interview?

No. Moxie is the interviewer, not a copilot. It never solves the problem or suggests code, and it doesn’t join Zoom or any other call. It’s for practice.

What can I practice on?

Anything you can put in a window: a LeetCode tab, your editor, a whiteboard, a design doc, a list of behavioral questions. You can also paste a problem or drop in a .md or .txt file.

Why do I need my own OpenAI API key?

So you pay OpenAI directly at cost, about $2.55 for a 45-minute interview, and nothing about your interviews passes through us. The key is stored in your Keychain.

What leaves my Mac?

Your microphone audio while an interview is live, a frame of your screen only when Moxie looks, and the problem text and transcript for prep and feedback. All of it goes to api.openai.com and nowhere else.

How long is an interview?

You choose. Each Kind has a default length, you can set a custom one up to 110 minutes, or turn the timer off and end when you’re done.

Windows or Linux?

Not today. Moxie is a native macOS app and needs macOS 15.2 or later.

Your next interview is out loud. Practice that way.

3 free interviews · no account needed