ReliAgent

ReliAgent intercepts and validates AI agent tool-call responses, running nine independent detectors against each call: sycophantic gap-filling, hallucination, context degradation, repetition loops, confidence collapse, parameter drift, timeout, schema violation, and response-contract integrity. When a supplied schema relies on a JSON Schema keyword outside the supported subset, ReliAgent returns an honest partial-validation finding instead of a silent pass. Returns a structured reliability report with per-detector confidence scores and remediation guidance.