Interviewers

An Interviewer is the AI persona that runs a Round. It tells the AI who to be and how to grade — the personality and the report card, bundled together. You don’t create an Interviewer on its own — you define it as part of creating a Round. The Round and its Interviewer are created together in one step (the second step of the Create Round form, or the nested interviewer object on the round API). This page explains every Interviewer field; see Rounds for how to send them.

What it controls

Who it is

The name and personality the candidate talks to.

How it talks

Friendly or strict, easy or hard, deep follow-ups or light ones.

How it scores

The list of things it grades, and how much each one counts.

Every field, in plain words

string

required

A label for you, like "Senior Backend Interviewer".

string

required

The first thing the AI says about itself, like "Hi, I'm Alex, a senior engineer." This sets the mood.

hiring | admissions | training | custom_api

required

Pick why you’re interviewing. This changes the words in the final verdict (for example hire vs ready to practice more).

array

required

The things to grade. Each one is { name, description, weight }. The weights must add up to 100. Example: System Design 50, Coding 30, Communication 20.

object

How the chat feels:

tone — friendly, professional, strict, or challenging
style — structured, conversational, or adaptive
difficulty — easy, medium, hard, or adaptive
probingDepth — low, medium, or high (how much it digs in)

string

Free-form directives for the AI in this round — the single AI-instruction field (focus areas, exclusions, closing questions).

object

How numbers are made: scoringMethod (weighted_average or rule_based). The score scale is no longer set here — new Interviewers default to a 0–10 scale.

object

Extra background so the AI sounds like it belongs: organizationContext, domainContext.

object

What ends up in the report: includeTranscript, includeScoreBreakdown, includeRecommendation, includeImprovementFeedback. All on by default.

How you make one

You define the Interviewer while creating a Round — there’s no separate create step:

Dashboard: open a Role, click Add round, fill in the round details (step 1), then the Interviewer step (step 2) — name, persona, use case, evaluation dimensions (weights add to 100), tone, difficulty, and AI instructions. Saving creates the Round and its Interviewer together.
API: send the fields below as the nested interviewer object on POST /api/v1/rounds. See the Rounds page for the full request.

To run a new Round on an interviewer another Round already uses, pass that round’s interviewer_id on the new Round instead of an interviewer object.

A filled-in example (no code)

Not technical? Here’s exactly what to type into each dashboard field to build a simple “Friendly Screener” — a warm first-round interviewer you can reuse for almost any job.

Field	What to type	Why
Name	`Friendly Screener`	Just a label for your list — candidates never see it.
Persona	`Hi, I'm Riya. I'll ask a few quick questions about your background.`	The AI’s opening line. Keep it to one friendly sentence.
Use case	`Hiring`	Sets the verdict wording (hire / no-hire).
Dimension 1	`Communication` — weight `40`	How clearly they explain themselves.
Dimension 2	`Motivation` — weight `30`	Why they want the role.
Dimension 3	`Relevant experience` — weight `30`	Do they have the basics?
Tone	`Friendly`	Warm, not intimidating.
Difficulty	`Easy`	It’s a first-round screen, not a deep-dive.

The three weights add up to 100 — that’s the only rule. Click Save, and your Friendly Screener is ready to drop into the first Round of any Role.

Each Round gets its own Interviewer, created with it. If you want the same Interviewer on another Round, pass that round’s interviewer_id on the new Round instead of describing a fresh one.

Examples for each use case

Hiring
Training
Mock interview

Strict and skill-focused.

{
  "name": "Backend Hiring Agent",
  "persona": "Alex, a staff engineer.",
  "useCase": "hiring",
  "evaluationDimensions": [
    { "name": "System Design", "description": "Scalable design", "weight": 50 },
    { "name": "Coding",        "description": "Correct & clean", "weight": 30 },
    { "name": "Communication", "description": "Clear trade-offs", "weight": 20 }
  ],
  "interaction": { "tone": "professional", "difficulty": "hard" }
}

Kind and coaching, so people keep learning.

{
  "name": "Sales Coach",
  "persona": "Sam, a friendly sales coach.",
  "useCase": "training",
  "evaluationDimensions": [
    { "name": "Discovery",         "description": "Finds the pain", "weight": 50 },
    { "name": "Objection Handling","description": "Handles pushback","weight": 50 }
  ],
  "interaction": { "tone": "friendly", "difficulty": "adaptive" }
}

Real-feeling but encouraging, with lots of feedback.

{
  "name": "Mock Interviewer",
  "persona": "Jordan, a big-tech interviewer.",
  "useCase": "training",
  "evaluationDimensions": [
    { "name": "Problem Solving", "description": "Approach & edge cases", "weight": 60 },
    { "name": "Communication",   "description": "Thinks out loud",       "weight": 40 }
  ],
  "interaction": { "tone": "challenging", "difficulty": "hard" },
  "output": { "includeImprovementFeedback": true }
}

Stuck on weights? Start with one big dimension (60) and one small (40), then adjust after a few interviews.

Roles

Put your Interviewer to work inside a Role.

Create a Round (API)

The interviewer is created as part of the round — see the endpoint reference.

Get Started

Concepts

Cookbook

Tutorials

Guides

MCP

What it controls

Who it is

How it talks

How it scores

Every field, in plain words

How you make one

A filled-in example (no code)

Examples for each use case

Next

Roles

Create a Round (API)

​What it controls

Who it is

How it talks

How it scores

​Every field, in plain words

​How you make one

​A filled-in example (no code)

​Examples for each use case

​Next

Roles

Create a Round (API)

What it controls

Every field, in plain words

How you make one

A filled-in example (no code)

Examples for each use case

Next