LANGGRAPH MULTI-AGENT SYSTEMS: Build Collaborative AI Agents, Supervisors, Handoffs, and Production Workflows with Python (The Practical LangGraph Series) (en Inglés)
Reseña del libro "LANGGRAPH MULTI-AGENT SYSTEMS: Build Collaborative AI Agents, Supervisors, Handoffs, and Production Workflows with Python (The Practical LangGraph Series) (en Inglés)"
BUILD RELIABLE MULTI-AGENT AI SYSTEMS WITH LANGGRAPH As AI applications grow, a single agent with too many tools, responsibilities, and instructions becomes difficult to control, test, and maintain. LangGraph Multi-Agent Systems shows you how to design collaborative AI systems where specialized agents work together through structured, testable workflows. Using Python and modern LangGraph patterns, you will learn how to build agent teams with clear roles, controlled context, defined permissions, and reliable coordination. Throughout the book, you will build IncidentPilot, a multi-agent incident-response system that combines application, database, security, communications, and supervisory agents. You will learn how to: • Build routers, supervisors, workers, and agent handoffs • Coordinate workflows with StateGraph, Command, Send, reducers, and subgraphs • Run agents in parallel using fan-out and fan-in patterns • Design structured state and communication contracts between agents • Control context, tools, permissions, and sensitive information • Add human-in-the-loop approval for high-risk actions • Prevent loops, duplicate work, uncontrolled delegation, and handoff failures • Test routing, supervision, parallel execution, and agent trajectories • Evaluate and observe multi-agent systems with LangSmith • Control latency, concurrency, model usage, and operational cost You will also build ChangeGuard, a collaborative change-review system that demonstrates parallel reviewers, structured findings, reconciliation, deterministic safeguards, and evaluation. By the end of the book, you will be able to design, build, test, and evaluate your own LangGraph multi-agent applications from the ground up. You will know how to choose the right coordination pattern, divide work between specialized agents, control what each agent can see and do, introduce human oversight where necessary, and surround model-driven decisions with deterministic safeguards. Just as importantly, you will understand when a multi-agent architecture is actually justified and when a simpler single-agent or deterministic workflow is the better engineering choice. Written for Python developers, AI engineers, and backend developers, LangGraph Multi-Agent Systems gives you a practical foundation for moving beyond experimental agent demos and building coordinated, testable, production-minded AI systems. Book 2 of The Practical LangGraph Series.
