Find AI vulnerabilities. Block them. Monitor them. Pick your plan.

One scan = one full run of the attack engine against one target, in any of the four scan modes (Browser / API / Agent / Model). A scan can include hundreds to thousands of individual attack techniques drawn from our 27,000+ library — the scan itself counts as 1 against your monthly quota.

A target is a single LLM endpoint you want to test + monitor — one chatbot URL, one API endpoint, one deployed agent, or one model slug. You can re-scan a target as often as you like (subject to your monthly scan quota).

Yes — every tier includes the runtime monitoring SDK (pip install shieldpi), the 6 detectors, the AI SOC analyst that runs every 60 seconds, and the CI bundle. Pro and above include higher event-ingest quotas and longer retention.

Yes — every scan produces deployable guardrails (Python middleware, LangChain callback, NeMo Colang, raw JSON) that you can download from the scan report or pipe into CI. No extra charge.

Yes. Enterprise and Team subscriptions accept USDC on Ethereum, Polygon, or Base. Contact sales to set up a billing address.

Scans above your quota queue for the next billing cycle — they don't fail, they just wait. If you need more headroom, upgrade a tier or contact sales about an add-on pack.

Yes — 14 days of Pro with no credit card required. Sign up, click the "Start Pro trial" button on this page, and you're in.

Enterprise customers can. We ship a dedicated single-tenant deployment (Kubernetes + Postgres + Redis) with your own attack corpus sync, your own Anthropic key, and our white-glove onboarding. Talk to sales.