Questions Founders Ask

Plain answers. No implementation jargon.

Questions

What does Agnostics test?

Agnostics runs adversarial attacks against your AI app: prompt injection, context leaks, tool misuse, retrieval drift, and other failure modes that happy-path testing misses.

What kinds of AI apps does it support?

Chatbots, RAG apps, AI agents, tool-using workflows, AI APIs, and custom AI products. If users can interact with it, Agnostics can pressure-test it.

Does Agnostics guarantee my app is safe?

No. Scans surface real failures with evidence. They cannot find every possible issue, and they do not replace your judgment about risk. Release Gate gives you a clear call, not a certificate.

What is a Release Gate?

Release Gate turns scan results into a ship decision: Ready, Monitor, Fix, or Blocked. It tells you what needs fixing, what can wait, and whether you should launch.

How do scans work?

You add a target, choose an attack pack matched to your app type, and run a scan. Agnostics sends adversarial prompts and checks responses against guardrails. Findings include severity, reproduction steps, and fix guidance.

Can I retest after fixing issues?

Yes. Retesting is core to the workflow. Fix a finding, run a retest, and watch your Release Gate update before you ship.

What happens to my target configuration?

Target details are stored in your account so you can rerun scans and retests. You control what you add and can delete targets when you no longer need them.

Can I cancel anytime?

Paid plans can be canceled through billing settings once accounts are live. Your access continues through the end of the billing period.

Who is Agnostics for?

Founders and teams shipping AI products who want release confidence without drowning in raw test output. If you are about to launch or iterate, Agnostics is for you.

Does it work before launch and after launch?

Yes. Run scans before your first release to catch breaks early. Keep scanning after launch when you ship changes, new tools, or updated models.