BranchLab

BranchLab is a local-first replay engine for AI agent runs. It lets engineers inspect a run, branch into counterfactual timelines, and review policy changes against evidence before those changes reach production.

• Counterfactual branching for agent timelines
• Local-first review without cloud upload
• Policy simulation before live deployment
• Evidence-grade run review for debugging and audit