Guardrails AI

Guardrails AI is a Python framework that adds runtime validation around LLM inputs and outputs. It wraps model calls with checks for hallucinations, PII, toxic language, format compliance, and policy violations, and can re-ask or fix outputs that fail.

Validators are pulled from the Guardrails Hub, a library of community and first-party checks, and composed into guards in code. It works across LLM providers and frameworks. The core is Apache 2.0, with a commercial offering layered on top for governance at scale.