Run screening interviews with your AI agent

First-round screens are repetitive, hard to schedule, and easy to run inconsistently. Your AI agent can run them for you — asking the same questions every time, listening carefully, and handing back a clean evaluation.

AgentCall is a skill you install into your existing AI agent — Claude Code, Cursor, Gemini, and more. It's not a notetaker or a standalone app. It gives the agent you already use a seat in the interview, so it can conduct the screen — or sit in and score it — and report back.

What an interview agent does

Two ways to run it

AI interviewer — the agent conducts the screen itself, asking your questions and following up. Silent evaluator — a human runs the interview and the agent sits in quietly, then returns a scorecard afterward. The prompts below cover both.

A note on fairness and consent

The bot is always a visible, named participant — never hidden. You write the questions; it runs them the same way for everyone; and the candidate knows they're speaking with (or being evaluated by) an AI. Use it for consistent first-round screens, with a human making the real decision.

Set it up in one sentence

"Install the AgentCall skill from github.com/pattern-ai-labs/agentcall and join the interview at <link> — create a free account for me if I don't have one."

Copy-paste prompts

Give your agent a name and tell it how to behave. Swap in your meeting link and questions, rename the bot, and paste one of these to your agent.

Prompt A — AI interviewer (conducts the screen)
Install the AgentCall skill from github.com/pattern-ai-labs/agentcall and join the interview at <meeting-link>. Create a free account for me if I don't have one. Join as "Riley", the screening interviewer. Conduct the interview: - Open with a brief, warm intro: you're an AI interviewer running the first-round screen, and the conversation is reviewed by a human afterward. - Ask these questions one at a time, in order: 1) ... 2) ... 3) ... Wait for the candidate to finish before responding. Ask at most one natural follow-up per question, then move on. - Never interrupt or talk over the candidate. Keep a warm, neutral tone — don't hint at "right" answers. - Keep it to about 15 minutes. - When it ends, return a structured evaluation: strengths, concerns, and a recommendation (advance / hold / pass) with a one-line rationale per question.
Prompt B — silent evaluator (never speaks; scores post-meeting)
Install the AgentCall skill from github.com/pattern-ai-labs/agentcall and join the interview at <meeting-link>. Create a free account for me if I don't have one. Join as "Quinn", a silent evaluator — a human interviewer is running this screen. - Do NOT speak in the meeting at all. Just listen and follow the full conversation. - Evaluate the candidate against these criteria: role knowledge, communication, problem-solving, [add your own]. - After the meeting ends, give me a structured scorecard: a rating per criterion, a few notable quotes, strengths, concerns, and an advance / hold / pass recommendation with a one-line rationale.

Works on every platform

The same skill on Google Meet, Zoom, and Microsoft Teams. New to this? Start with how to make your AI agent join a meeting.

Frequently asked

Does the candidate know it's an AI?

Yes. The bot joins under a name you choose and introduces itself. It's never a hidden participant.

Can it just sit in and score, without conducting?

Yes — use Prompt B above. It joins silently while a human runs the interview, then returns a structured scorecard.

What does it cost?

6 free hours, then from $0.35/hr. See pricing.

Run your next screen with an AI interviewer.

Start free — no card needed