Real-time safety for LLMs
Guardrail API is a firewall for your AI stack.
Guardrail sits between your users and LLMs, intercepting prompts and responses in real time. Enforce policy packs, catch prompt injections, and ship safer AI features without rebuilding your stack from scratch.
Prefer to read first? Open the documentation.
Core Guardrail API
Drop Guardrail in front of any LLM. Send us prompts and responses; we handle policy checks, sanitization, and decisioning.
- • Ingress and egress policy enforcement
- • Low-latency decision endpoints
- • JSON-native events for your logs and SIEM
Enterprise Admin Console
Give operators a single console to manage tenants, monitor traffic, and respond to incidents without touching YAML.
- • Tenant and bot management with RBAC
- • Live traffic views and incident timelines
- • Usage, billing, and audit-friendly exports
Policy packs & overlays
Start with built-in safety packs, then extend them to match your own compliance posture.
- • Pack-based rule configuration
- • GDPR / HIPAA style overlays
- • Versioned policies with change history
Verifier integration
When rules aren't enough, delegate gray-area decisions to a verifier model without ever executing untrusted code.
- • Verifier request/response contracts
- • Fallback chains across multiple LLMs
- • Token budgets and timeout controls