Skip to content

Demo

Watch the agent work, step by step

These are real recorded episodes, scored against ground truth. Pick one and watch every tool call the agent made on the way to its answer. Running a live model takes your own API key and the project run locally, from the source.

45 recorded episodes · bench v1.0

Measured 2026-08-20 on bench v1 (8 perturbations, 9-type flag vocabulary) with claude-sonnet-4-6. Case inputs rebuild byte-identically under current code; aggregate metrics from this archive must NOT be pooled with bench v2 results, whose prompt vocabulary differs.