Setup

This page walks you through everything from downloading SyntaxCue to your first hint — install, entering your API key, granting permissions, and starting to listen. Worth doing once calmly before a real interview, rather than the moment one starts.

1. Download

Get SyntaxCue from the Download page. It’s free to try — no purchase required upfront. You get around 10-15 free hints (metered by hint count, not a countdown clock) before you need a license.

Builds are published on GitHub Releases, not a custom update server. Each release includes a SHA-256 checksum so you can verify your download — see the Security page for how.

2. Install

macOS

1. Open the downloaded .dmg.
2. Drag SyntaxCue into your Applications folder.
3. Launch it from Applications or Launchpad.

The macOS build is code-signed and notarized by Apple, so on first launch you’ll see the standard “downloaded from the internet — Apple checked it for malware” confirmation, not an “unidentified developer” block. Click to open and you’re in. (Launch it from Applications rather than running it from inside the mounted .dmg volume.)

Windows

1. Run the downloaded installer (-setup.exe).
2. Windows SmartScreen will show “Windows protected your PC” on first run. This is expected: the Windows build isn’t code-signed yet, so SmartScreen doesn’t recognize it — not a sign anything’s wrong with your specific download.
3. Click “More info”, then “Run anyway” to continue.

If you want to confirm the file wasn’t tampered with before running it, every GitHub release posts a SHA-256 checksum you can compare against — the exact commands are on the Security page. You don’t need to disable any antivirus or security software to install.

3. First launch — the setup screen

When SyntaxCue opens, the setup screen asks for a few things:

1. LLM provider — a dropdown with two choices: OpenAI or Anthropic (Claude).

2. API key — paste in your key. It’s stored locally in your OS’s secure keychain (macOS Keychain / Windows Credential Manager) and never sent anywhere except straight to your chosen provider’s API. There’s a collapsible “How do I get an API key?” section right in the app that walks through creating one from the provider’s dashboard.

   This is the one real prerequisite: you need an OpenAI or Anthropic API account with billing enabled and some credit on it. SyntaxCue is bring-your-own-key, so your usage is billed to you directly by the provider. If your first hint fails, an empty or unbilled API account is the most common reason.

3. Model — a free-text field where you type the exact model name you want to use. The app’s own note: “Must support vision to answer screenshots. Defaults to a fast, current model for the selected provider.” It’s an open field on purpose, so it doesn’t go stale as providers release new models — if you want to answer coding-problem screenshots, pick a vision-capable model.

4. Preferred language / framework (required) — type what you’re interviewing in, e.g. “Go”, “Python/Django”, “TypeScript/React”. Code answers use this instead of defaulting to Python.

5. Answer style — Short, Balanced, or Detailed. Live coding rounds often want Short; a system-design discussion might want more.

6. Interview language and, optionally, a separate response language — the first is for speech recognition, the second for the language the model replies in. Most people leave both the same; set them independently only if you want the reply in a different language than the interview.

4. Permissions

SyntaxCue has a dedicated Permissions section with “Check / enable” and “Allow” buttons. Walk through it now, before an interview, rather than during one:

• System audio access (required) — this is how SyntaxCue hears the interview at all. Without it, there’s nothing to transcribe.
• Screen Recording access (macOS only, and only if you enable the optional region-screenshot hotkey) — needed for that one feature only. If you’re not using the hotkey, you don’t need this.

5. Start listening

Click “Start listening” and wait for the “Listening” status to appear before you speak. Both the whisper.cpp model and audio capture take a few real seconds to initialize on start. Speaking too early looks exactly like a capture bug from your side, so give it that moment. Once it says Listening, you’re live.

Adding a coding problem

When there’s a coding problem on screen, you send it to the model — a few ways, depending on the situation:

• The problem is visible on screen right now — click the “Screenshot” button to capture the whole screen instantly. If you’ve set a hotkey for it in Advanced settings (off by default), pressing it does the same thing without touching the mouse.
• The problem is long and scrolls, so no single screenshot shows all of it — select its text instead (⌘A / Ctrl+A to select all, if it’s the only thing on the page), copy it, and paste it in (⌘V / Ctrl+V). Pasted text works exactly like a screenshot and doesn’t need a vision-capable model.
• You already have a screenshot — from your OS’s own shortcut or a third-party tool — paste it in (⌘V / Ctrl+V) the same way.

The model reads the screenshot image (or pasted text) and the audio transcript together. There’s no continuous screen reading, OCR loop, or IDE integration — each capture is an explicit screenshot or paste you trigger.

For how to verify your download and exactly what leaves your machine, see Security. Once your free hints run out, a one-time $79 license unlocks unlimited use for 6 months — details on Pricing.