Guardrails Concepts

Guardrails make agent behavior safer and more predictable by validating inputs/outputs and enforcing policies (e.g., PII detection, content constraints, prompt-injection resistance).

Time to read: ~5 minutes.

How Idun applies guardrails

• Consistent enforcement across supported frameworks (LangGraph, ADK, Haystack)
• Config-driven: policies are attached via configuration and can be managed centrally in managed mode
• Auditable: tie guardrail decisions to traces and logs for compliance and debugging

Next steps

• Guardrails overview
• Configuration concepts