2.0

Guardrails are now available to help keep your agents and their inputs secure. These built-in safeguards protect against PII leaks, prompt injections, jailbreaks, and NSFW content.

Add guardrails with a single line of code by importing one and passing it as a pre_hook when creating your agent.

You can also extend the BaseGuardrail class to build custom safeguards tailored to your needs.

Learn more in the documentation.