Try it now
Find the root cause of an agent failure
Paste a trace from your agent run. ProveAI Origin compresses it, runs the judge, and points at the span that caused the failure — with cited evidence and a failure category. No sign-up.
The longer an agent runs, the more a single wrong step costs. Load a sample — a simple CX agent, a long-running coding agent, or a frontier research run — and see what the failure actually wasted.
Your trace stays private. We process it in-memory to run the attribution and never write the raw trace to our database. Only the attribution result is kept, for 24 hours, to power the share link — then it's deleted. Signing up later does not link this trace to your account.