All projects matching “guardrail”

DCL Trust Oracle. Call only when the text must include the substring AI and must not include "I cannot be held responsible" or "no guarantees". POST JSON {"response","agent_id"}. COMMIT means those checks passed and the action may proceed. NO_COMMIT means do not execute the checked action. Not a tool-call guardrail, prompt-injection detector, or jailbreak screen; use /evaluate/fast for the phrase

Agent guardrail decision engine. Call once before every agent action; returns a single decision — proceed, slow down (with a reason and wait time), or stop — based on cost budget, chain depth, step cap and retry budget, and request volume pressure. Policy thresholds are configured separately via /policy; this endpoint only reports the decision for a single action.

Overwing guardrail: check text before you send, post, or act on it. Content moderation, PII detection, confidential leak detection, toxicity, self-harm and severity in ~150 ms. Returns a verdict plus one recommended_action (block / redact / review / allow) with calibrated confidence. Pass context (recipient, channel, owns_contact_info) with the outbound-message set. $0.002 USDC, no account.

For model cascades: is the small model's answer usable, or should the larger model be paid? Detects refusal, evasion, off-topic answers, contradiction and incompleteness; decides accept, review or escalate. Does not fact-check without a source: provide the source to check substance.