For one team getting its first LLM feature through review.
- 1 production app
- 1M Guard requests / month
- 2,000 eval runs / month
- Monthly automated red-team scan
- Community + email support
Priced on protected requests and eval runs, not seats. Every plan includes Red Team, Evals, Guard and Trace.
For one team getting its first LLM feature through review.
For companies with several AI features and a security team that asks questions.
For regulated companies and anyone who needs it inside their own walls.
| Capability | Starter | Growth | Enterprise |
|---|---|---|---|
| Probe Guard (input & output policies) | ■ | ■ | ■ |
| PII & secret redaction | ■ | ■ | ■ |
| Eval runs | 2,000 / mo | Unlimited | Unlimited |
| Red-team cadence | Monthly | Weekly | Continuous |
| Agent & tool-call attacks | — | ■ | ■ |
| Trace retention | 14 days | 90 days | Custom |
| SSO / SCIM | — | SSO | SSO + SCIM |
| Self-hosted deployment | — | — | ■ |
| Compliance evidence packs | — | SOC 2 | SOC 2 · ISO 42001 · EU AI Act · HIPAA |
| Support | 8h SLA | 30 min, 24/7 + named engineer |
One model call that passes through Guard, regardless of token count. Streaming responses count once. Requests blocked before reaching the model still count.
No. You bring your own provider keys and pay your provider directly. We never resell tokens.
Yes. Every demo ends with a 14-day pilot on your own app, including a full red-team report. No card required.
Nothing breaks. Guard keeps protecting traffic and we bill overage at your plan's per-million rate. We'll warn you at 80%.
Yes. Companies under 30 people or with less than $5M raised get 50% off Growth for the first year.