LLM observability and evaluation platform that traces every step of AI agent workflows, tracks costs and latency, and connects findings to concrete improvements.