ASSERT

Open-source, framework-agnostic evaluation harness for AI agents and LLM applications that turns natural-language behavior requirements into executable evaluations, generates test scenarios, runs targets, and scores traces with evidence.