Skip to main content
The API provides three evaluation endpoints, one per policy mode:
EndpointModeRequest bodyUse case
POST /api/policies//evaluate/conditionsConditionsrequest.input (object)Structured rules (amount, user, resource).
POST /api/policies//evaluate/instructionsInstructionsrequest.prompt, request.input (string)Prompt + LLM response validation.
POST /api/policies//evaluate/guardrailsGuardrailsrequest.input (string)Text safety, PII, moderation.
All three return the same response shape: an envelope with meta (code, message, errors) and result (violated, decision, reason, key).

Evaluate conditions

POST /api/policies/:key/evaluate/conditions

Evaluate instructions

POST /api/policies/:key/evaluate/instructions

Evaluate guardrails

POST /api/policies/:key/evaluate/guardrails