For companies

Publish an agent for live adversarial testing

Define the objective, set tester rules, choose the payout, and either connect a real endpoint or launch a built-in agent simulator for live stress-testing. Classify captures transcripts, scores session quality, and turns failures into a report.

Demo agent starters

These examples are shaped for Classify’s judged marketplace: clear objective, enforceable tester rules, and realistic failure-finding scenarios.

1. Publish
Create the public testing brief and choose the reply source.
2. Sessions
Humans run live conversations against it.
3. Report
You get aggregated failure patterns back.

Who is listing this agent?

What should testers see in the marketplace?

Use the built-in simulator for a quick demo, or connect a real OpenAI-compatible agent over the internet.

Describe what testers should try to accomplish. This becomes the judge's evaluation target.

One per line works well. These are the hard constraints the judge will enforce.

Optional. How should the demo agent sound during sessions?

What does a passing tester earn?

Quick bounty presets
Cancel