模擬面接

An AI mock interviewer
that reads your screen.

Code in your own editor. Mogi watches, listens, and pushes back — the way a real interviewer would. No problem bank, no copy-pasting, no switching tabs.

Download for macOS

Apple Silicon & Intel

The screenshot is the problem.

Mogi has no question text of its own. It floats over whatever you are already using — VS Code, IntelliJ, a terminal, LeetCode in a browser — and reads both the problem and your code straight off the screen. You never paste anything in.

  1. 1

    Open a problem anywhere

    Your editor, your setup, your keybindings. Mogi adapts to you rather than the other way around.

  2. 2

    Talk it through

    Tap the voice hotkey and explain your approach out loud — exactly what a real interview actually tests.

  3. 3

    Get pushed, not rescued

    Hints, counterexamples, "what's the complexity there?" Mogi is built to never hand over the solution.

What you get

Company practice

Draw from real question pools by company, weighted by how often each problem actually shows up.

Voice, both ways

Speak your reasoning and hear the interviewer reply. A global hotkey works even while your editor has focus.

Transcripts & debriefs

Every session is saved locally with a full transcript and a written debrief of how it went.

Stays on your machine

Sessions live in a local database. Your API keys never leave the app, and there is no telemetry.

Installing

  1. Download the .zip above and unzip it.
  2. Drag Mogi.app into your Applications folder.
  3. Right-click the app → Open → Open.

    Mogi isn't code-signed yet, so macOS blocks it on the first launch with "Mogi is damaged and can't be opened." It isn't damaged — right-click → Open adds a permanent exception. You only do this once.

    Prefer the terminal? xattr -cr /Applications/Mogi.app

Bringing your own keys

Mogi runs on your own API keys — you pay the providers directly, at cost, with no markup. Paste them into Settings on first launch; they're stored on your machine and never sent anywhere but the provider.

OpenRouter

Required

Runs the interviewer itself — reading your screen and asking the questions. Pay-as-you-go, and a practice session costs cents.

How to create one
  1. Sign in at openrouter.ai (Google or GitHub sign-in works).
  2. Add a few dollars of credit under Credits. OpenRouter is prepaid — with a zero balance every request fails.
  3. Go to KeysCreate Key, and give it a name like Mogi.
  4. Copy it — it starts with sk-or- and is shown only once.
  5. In Mogi: Settings → API Keys → OpenRouter, paste, save.

Google Cloud

Optional · voice

The low-cost path for spoken interviews. Fiddlier to set up than the other two, because Google needs two separate APIs switched on.

How to create one
  1. Open the Google Cloud console and create a project (any name).
  2. Enable Cloud Text-to-Speech — this is the voice you hear.
  3. Enable Cloud Speech-to-Text — this is your microphone. Both are needed; enabling only the first leaves the mic failing silently.
  4. Go to CredentialsCreate credentialsAPI key. It starts with AIza.
  5. Optional but worth it: click Edit API key → restrict it to just those two APIs, so a leaked key can't be used for anything else.
  6. In Mogi: Settings → API Keys → Google Cloud, paste, save.

The project needs billing enabled even though both APIs include a free monthly allowance — Google won't turn them on otherwise.

ElevenLabs

Optional · voice

The premium path — Scribe for listening, Flash for speaking. The most natural-sounding option, and the quickest to set up.

How to create one
  1. Sign up at elevenlabs.io. The free tier is enough to try it out.
  2. Open Settings → API Keys and create a key.
  3. Copy it — it starts with sk-el- and is shown only once.
  4. In Mogi: Settings → API Keys → ElevenLabs, paste, save.

Voice is entirely optional — Mogi works typed. Either voice key on its own gives you the full experience; add both and Mogi automatically pairs ElevenLabs for listening with Google for speaking, which is the best combination of the two.