Find what your AI agent gets wrong — before you have a rubric. Qualitative eval for PMs.

Local-first LLM evaluation runner with baselines, caching, markdown reports, and CI-friendly quality, latency, and cost gates.

Binary safety verdicts (SAFE/HELD/LEAK/MISS/BROKE) + persona fan-out for LLM pipeline evals

Open-source evaluation framework for AI agents. Define test suites with rubrics, run your agent, get LLM-as-judge scores against criteria, inspect full execution traces, and diff runs to catch behavioral regressions.