Glossary

Guardrails (AI Agent)

**Guardrails (AI Agent)** are configurable safety controls that enforce predefined policies on AI agents by scanning inputs, tool calls, tool responses, and outputs to detect and mitigate risks like harmful content, prompt injections, or policy violations in real-time.[1][2][4]

**Guardrails (AI Agent)** are configurable safety controls that enforce predefined policies on AI agents by scanning inputs, tool calls, tool responses, and outputs to detect and mitigate risks like harmful content, prompt injections, or policy violations in real-time.[1][2][4]