Bring Your Own LLM Key — How and Why Interview Cheat Uses Your Key

What "bring your own key" means in practice

Interview Cheat asks you to add your own Anthropic API key (for Claude models) or your own OpenAI API key (for GPT models) in Settings the first time you launch the app. Every interview turn — the transcribed question plus your resume plus the system prompt — is sent directly from your Mac to Anthropic or OpenAI using your key. The drafted answer streams back to your overlay. Interview Cheat's servers are not in the loop.

You pay your provider directly. Anthropic and OpenAI bill you for the tokens you used; we do not see that bill or add anything to it. A typical 45-minute interview costs $0.50 to $2.00 in API usage on current Claude or GPT pricing.

Reason 1 — Latest models the day they ship

When Anthropic or OpenAI ships a new model, you can use it the same day by changing the model name in Settings. You do not wait for Interview Cheat to update, integrate, and re-deploy a new build with the new model wired in.

This matters more than it sounds. Senior interview rounds are calibrated against the current state of model output. A vendor product running last-quarter's model is noticeably behind a product running this week's model. With BYO key, you control the model selection on your own clock.

Reason 2 — Cost, with no markup

Hosted interview-copilot products either charge a subscription that includes inference cost — which means they need to mark inference up to cover overage from heavy users — or charge per-minute / per-interview with similar margin built in.

BYO key removes the markup. You pay Anthropic or OpenAI at exactly their list price. If you use the app for one interview, you spend a dollar or two on tokens. If you use it for thirty interviews across a job search, you spend thirty to sixty dollars. Plus the $29.98 one-time Invisible Mode unlock, the total cost is typically well under the equivalent two-month subscription on a competing tool.

Reason 3 — Your interview content never touches our servers

Privacy is the reason BYO is the only design we considered. Interview transcripts can include compensation discussion, references to specific people, ongoing job-search status, and details about projects under NDA. Holding that data on Interview Cheat's servers would create a category of risk we do not want — for you or for us.

By streaming directly from your Mac to your chosen LLM provider, the data lives only in two places: your machine and the provider you chose. Anthropic and OpenAI both have published data-retention and training policies you can verify. Interview Cheat is not in the data path.

The tradeoff — you have to set it up once

The cost of BYO is the first-launch friction. You have to go to console.anthropic.com or platform.openai.com, generate a key, and paste it into Settings. For a non-technical user this is a non-trivial five-minute setup.

A future "Pro" tier with hosted models is on the roadmap for users who would rather not manage a key. That tier will be subscription-based by necessity. The free tier with BYO is staying free; the hosted Pro tier is the future paid option for convenience.

Technical details

Your API key is stored locally on your Mac in the macOS Keychain. It never leaves your machine except as the Authorization header on requests to api.anthropic.com or api.openai.com.

You can rotate the key at any time. Revoke the old key in the provider console; paste the new one into Settings. There is no Interview Cheat-side state to update.

Switching providers is one Settings field. If you have keys for both, you can swap mid-job-search without losing any resume or framework state.

Further reading

• Anthropic — Claude API documentation
• Apple — ScreenCaptureKit framework documentation
• Interview Cheat — open-source desktop app on GitHub