All features

AI is how people code now. Assess for it, do not pretend it away.

Candidates get a budget-capped AI assistant right inside the workspace. Every prompt is recorded, so you measure how well they direct it - not whether they managed to sneak it in.

Start freeSee pricing

Screenshot coming soon

Use cases

When this feature earns its keep.

Three concrete patterns. Use them as templates for your own setup.

Assess the skill people actually use

Banning AI tests for a world that no longer exists. Letting candidates use it, on the record, shows you who decomposes a problem and directs the model well versus who pastes the prompt and hopes.

Bounded and fair by design

Each assessment carries a hard AI budget. The assistant stops when it is spent, so there are no runaway costs and every candidate works within the same allowance.

Nothing happens off the record

Every prompt the candidate sends and every reply comes back in the reviewer transcript, machine-context noise stripped out. You read the actual exchange, not a summary.

Setup

How the setup looks.

The assistant is a per-assessment switch. When it is on, the candidate runs Claude Code in their workspace terminal and you get the transcript on review.

Same workspace, same auth, same collaboration features as every other Typelets workspace. See all features

  1. Enable the assistant for the assessment

    Turn the AI assistant on in the assessment settings. Leave it off for a no-AI assessment, on for a realistic one.

  2. The candidate pairs with it while they work

    They run Claude Code from the workspace terminal - asking questions, generating code, and iterating - the same agentic CLI they would use day to day.

  3. You read the transcript on review

    Open the AI transcript for the submission to see every prompt and reply in order, with each candidate identified. Toggle a raw view to audit exactly what was sent.

FAQ

Does the assistant just hand the candidate the answer?
It is a general coding assistant, not an answer key, and the whole point is what the candidate does with it. A strong candidate directs it well and verifies the output; a weak one delegates wholesale, and the transcript makes that obvious.
How is the budget controlled?
Each assessment has a hard spend cap enforced server-side. Once it is reached the assistant stops responding, so a candidate cannot run up cost and everyone gets the same allowance.
Is the assistant required?
No. It is enabled per assessment. Turn it off entirely for a no-AI take-home, or on when you want to assess how the candidate works with AI.

Start interviewing in under a minute.

No card, no install, no yaml. Sign up, create a workspace, share the link. That's it.

Free during beta · No credit card · Cancel anytime