Interviews overview

Typelets has a first-class mode for technical interviews: a shared coding workspace where an interviewer poses a problem, the candidate works in real time, the session is recorded, and the result is scored against a rubric.

The shape of an interview

1. Create an interview workspace. Set its mode to interview. Interview workspaces cannot be public.
2. Apply a problem from your problem library - this lays down starter files and attaches the prompt, rubric, and test cases.
3. Add the candidate as a member with the candidate role, and yourself (or colleagues) as interviewer.
4. Start the interview. Optionally start a recording.
5. The candidate codes; you collaborate, watch the shared terminal, and run tests.
6. End the interview - the candidate loses access immediately.
7. Score the candidate against the rubric, and review the recording with your team.

What candidates can and cannot see

Candidates see the prompt, their own files, and public test cases. They never see the rubric, hidden test cases (beyond pass/fail), the reference solution, or scores. See Sharing & roles.

Roles in an interview

• interviewer / admin / owner drive the session: apply problems, start/end, record, and score.
• candidate participates and is locked out when the interview ends.

Driving interviews programmatically

The MCP server exposes interviewer tools - apply a problem, summarize or score a recording, suggest follow-up questions - so an AI assistant can help run or review a session.

Next: Problem library.