Why Multi-Agent Systems Fail: The MAST Taxonomy

This chapter explains "Why Do Multi-Agent LLM Systems Fail?" by Anna Alexandra Grigoryan, which summarizes a research paper that set out to study failure systematically. Where Chapter 3 argued from principles that multi-agent systems are fragile, this chapter brings evidence: a careful, data-driven catalog of how they actually break, and a method for diagnosing your own system.