Multi-Agent Systems: A Contemporary Treatment

A textbook that teaches the concepts and techniques of Multi-Agent Systems in the era of LLMs.
Authors

Dell Zhang (XSci Ltd)

Jun Wang (University College London)

Benjamin Chang (University of Cambridge)

Working draft of

6 September 2026

About This Book

Book cover of Multi-Agent Systems: A Contemporary Treatment by Dell Zhang, Jun Wang, and Benjamin Chang. Beneath the title, the outline of a human brain is drawn as a network of interconnected nodes, with small robot, drone, and chart icons at the junctions; the words Working Draft and the book's web address stand at the foot.

Who it is for. Engineers building agent systems who suspect those systems are re-enacting a theory nobody showed them; students of multi-agent systems who want the classical results run on machines that finally exist; and anyone who must decide whether to build several agents at all.

What it teaches. Three things above all: how to price the decision to use more than one agent, and when a single agent or a fixed workflow wins; which classical results — on communication, cooperation, coordination, strategic interaction, collective choice, and institutions — transfer to language-model agents, and on what assumptions; and how to build, evaluate, secure, and account for a team of agents, from a runtime written from scratch to a framework you can leave.

Start reading Run the first example Download the PDF Download the EPUB

NoteCite this book 
@book{zhang2026masact,
  title = {Multi-Agent Systems: A Contemporary Treatment},
  author = {Zhang, Dell and Wang, Jun and Chang, Benjamin},
  year = 2026,
  note = {Manuscript in progress, written in public},
  url = {https://books.bloo-mind.ai/masact/}
}