Last released Nov 19, 2026
Unit tests and a CI gate for LLM agents: reliability (pass^k), multi-turn conversations, prompt injection, hallucination, loops, cost and latency. Works with OpenAI Agents SDK and LangGraph.