video1.0<iframe src="https://www.loom.com/embed/6d0e715eae75445092a97d69862c55b0" frameborder="0" width="2094" height="1570" webkitallowfullscreen mozallowfullscreen allowfullscreen></iframe>15702094Loomhttps://www.loom.com15702094https://cdn.loom.com/sessions/thumbnails/6d0e715eae75445092a97d69862c55b0-25eff695badc74b9.gif300.683This Loom presents AgentHabitat, an open source AI agent verification tool aimed at regulated industries where outputs must be provable. It explains a verification layer that checks every factual claim word by word against original sources, stops if claims cannot be traced, and logs every AI call and decision for searchable auditability. It also preserves human approval states so work can resume after a crash, and only runs expensive AI calls after sign off. In a live demo, the verification prevented an ungrounded claim, triggered a retry, and completed the task at under $0.13, specifically $0.12, with grounded claims verified against the source.