Percival is a highly intelligent agent developed by the Patronus AI team. It is capable of detecting 20+ failure modes in agentic traces and suggesting optimizations for agentic systems. Think of Percival as your best AI debugger who has spent thousands of hours understanding your traces and processing millions of tokens.

We’ve seen firsthand how many hours AI engineers spend combing through agent traces searching for agent planning mistakes, incorrect tool use, and context misunderstanding. We built Percival to make agent debugging significantly faster: with the click of a button, Percival surfaces 20+ failure modes and suggests improvements to fix them.

Traditional evaluation approaches like LLM-as-a-Judge or static evaluation runs can catch mistakes at specific points, but often miss the broader context and overlook systemic issues like flawed planning.

And Percival gets better over time. Each time you confirm an issue or annotate a missed failure mode, Percival learns, helping you make your evaluation domain-specific and build confidence in your agentic AI.