Jevyr

Jevyr is a local-first AI judgement system that tests AI-generated claims, code, plans, and decisions by running controlled experiments and collecting evidence to produce accepted, rejected, or unproven results. It provides a framework with a local server, CLI, SDKs, and a human-inspectable interface to compare explanations and verify AI outputs with replayable evidence records. The project is experimental, complex, open source, and intended for forking and further development.