Interview eval template

Candidate and evaluators share one live room. AI analysis stays on role:evaluator messages.

Setup

1. Create room interview-{candidateId}.
2. Add candidate as member or guest.
3. Add evaluators and AI agent with room_members.role = evaluator.

Agent posts evaluator-only notes

await client.createMessage(roomId, analysis, {
  visibility: "role:evaluator",
});

Candidate view

Candidate JWT timeline shows public questions only. Evaluator dashboard shows public plus role-scoped notes.

Docs

Role-scoped visibility · Room as MCP server for external Claude evaluator.