Runtime Monitor

The subsystem that evaluates every live response, classifies failures, and triggers self-healing or escalation.

The Runtime Monitor watches every response from every active connector — no sampling. It does not merely check for HTTP errors; it classifies each failure into a taxonomy: transient network fault, schema change, API version change, authentication failure, rate limiting, or unclassified.

Each class has a defined response. Transient faults retry with backoff. Schema changes trigger the self-healing loop. Auth failures suspend the connector immediately and alert a human, because credentials are managed externally and must never be retried blindly.

This is what turns troubleshooting into monitoring.

See it in action