Automated pen-testing

Pick an agent · launch 30 attacks · get a resilience score

Click any agent below. SENTRY runs the OWASP LLM Top 10 attacks, classic jailbreaks (DAN, AIM, grandma), indirect injection, PII fishing, EU AI Act traps, DORA incident denial, and authority impersonation against it. Returns a 0–100 resilience score plus the exact prompts that broke through.

attacks9 categories~120 s runtime0–100 score

Pick a target agent

Click any of the agents SENTRY is currently auditing. The pen test runs the full attack suite against the selected agent — no API keys required for the built-in demo agents.

Loading agents…
⚙ Use a custom external agent instead (advanced)