designing-multiagent-systems

Why we included this project

This is the companion codebase to Victor Dibia's book on designing multi-agent systems, and it works best as a structured learning path rather than something you'd drop straight into production. The chapters move from first principles to complete, tested implementations, anchored by PicoAgents, a deliberately minimal framework built from scratch so every design decision stays visible. Along the way you get a first agent with tools and memory, computer-use and browser automation, type-safe workflows, orchestration patterns like group chat and plan-based control, LLM-as-judge evaluation, and a production case study that analyzes thousands of companies. The framework-agnostic explanations and side-by-side comparisons of Microsoft Agent Framework, Google ADK, and LangGraph make it a good fit for developers and tech leads who want to understand how multi-agent systems actually work before committing to one. Treat it as a thorough tutorial and reference, not as a framework to deploy directly.