The open frameworks for enforcing policy at the prompt, output and tool-call layer.
Guardrails AI and NeMo Guardrails are the leading open frameworks for enforcing policy on LLM inputs, outputs and agent actions. They provide programmable validators, structured output, topical guards and tool-call filters, executed deterministically alongside the model.
I design guardrail policy sets for your agent estate, integrate the frameworks into your runtime, write custom validators for your domain, and instrument the observability that proves the guardrails are firing as intended.