retrace
RecordOne decorator captures every LLM callFork & ReplayBranch from any decision, replay the futureProve the FixRe-run a fix and get a verdict — did it work?Eval GatesCI/CD quality gates that block bad deploysGuardrailsRuntime safety — halt agents before they failEnforcementCircuit breakers — block runaway loops & budget blow-outsFailure TaxonomyMAST — why agent runs fail, auto-classifiedSessionsMulti-agent tracing with causal graphs
Pricing
Playground
Docs
Changelog
Sign inGet Started

Explore Tapes

Watch AI agents think, step by step.

prod-flight-agent

✓ completed⚠ anomaly
70