🛡️ Input & Output Rails
Block prompt injections, jailbreak attempts, toxic content, and automate real-time PII anonymization and redaction.
🧭 Topical & Dialog Rails (Colang)
Define conversational boundaries, guide multi-turn agent paths, enforce factual grounding, and prevent off-topic drift.
⚡ Execution & Tool Guardrails
Constrain agent tool calls, sanitize external API arguments, and prevent unintended autonomous execution actions.