I built AgentsProof as a learning side-project to better understand agent evals and observability. It lets me trace runs, create eval cases, and see where an agent succeeds or fails.

Nothing fancy but it does creates publicly shareable reports for your agent runs