Interview Service

Metrika · Interview Service

AI-led interviews from a single API call

Define the goal and the checklist, get back a link — an AI interviewer conducts the conversation. Watch and steer it live while it runs, and get a structured, machine-readable result at the end.

Invited to an interview? Open the personal link from your invitation — there is nothing to do on this page.

How it works

  1. Create it

    One POST /v1/interviews call with the goal, the checklist and the rules. The response carries the respondentUrl.

  2. Deliver the link

    You send it yourself — by e-mail, chat, ticket. The service never contacts anyone on its own.

  3. The AI interviews

    A 10–20 minute conversation on the hosted page. Watch it live (SSE) and steer with notes when you learn something mid-flight.

  4. Structured result

    On close, the transcript is harvested into your resultSchema, and a signed webhook tells you it is ready.

What it does

Interviews and campaigns

One private link for one person — or one shareable campaign link that mints a separate private interview for every opener.

Agents

A tested interview archetype stored once; a launch fills only the open slots. Every launch freezes its own copy, so a mid-wave edit can never corrupt comparability.

Live observation and steering

An SSE stream of every event, and guidance notes while it runs. The interviewer takes them into account — it does not read them out.

Structured results

The transcript harvested into your own resultSchema; HMAC-signed webhooks tell you when the result is ready.

MCP server

21 tools at /api/mcp, mirroring the REST API 1:1. Claude Code, Claude Desktop and claude.ai connect with an API key or OAuth.

Hosted respondent page

Mobile-friendly chat with autosave and resume, no sign-in: the link is the credential. With a privacy notice and consent built in.

Want to try it?

The service is used with an API key; keys are currently issued by Metrika. With a key in hand, the first interview is a single curl:

curl -s -X POST https://interview.getmetrika.hu/v1/interviews \
  -H "Authorization: Bearer $KEY" \
  -H 'Content-Type: application/json' \
  -d '{
    "config": {
      "goal": "Understand how this person uses the product day to day.",
      "checklist": [
        {"id": "USAGE", "label": "Usage", "description": "What they do with it, how often."},
        {"id": "FRICTION", "label": "Friction", "description": "Where it gets in their way.", "required": true}
      ],
      "language": {"start": "hu"}
    }
  }'

Or put an AI agent on it: over the MCP server, Claude creates, watches and harvests interviews itself —

claude mcp add --transport http interview https://interview.getmetrika.hu/api/mcp \
  --header "Authorization: Bearer ivk_..."

The complete reference — config schema, webhooks, errors, recipes — is in the API documentation.