Multi-agent system
A multi-agent system is a computational system composed of multiple agents that interact or coordinate to perform tasks or pursue goals.
Introduction to Generative AI, Second Edition
A multi-agent architecture is a system design in which multiple software agents coordinate or interact to perform tasks.
- How it is used
- Coordinates specialized AI agents to divide complex workflows and pursue shared goals.
- How deep it goes
- Defines roles, communication, delegation, and synchronization at a high level, without implementation instructions.
- Examples
- Illustrates employee onboarding, coordinated scientific research, and drug-discovery workflows.
- What you can do after
- Evaluate/understand it as a coordinated workflow whose group behavior and risks require system-level assessment.
Read how Introduction to Generative AI, Second Edition explains multi-agent system →
AI Agents in Action
A multi-agent system is a system in which multiple specialized agents work and communicate together to solve problems.
- How it is used
- Coordinates specialized agents to divide complex work, run tasks in parallel, and provide internal feedback or evaluation.
- How deep it goes
- Explains profiles, proxies, communication patterns, setup, execution, and exercises, while deferring deeper platform implementation details.
- Examples
- Uses coder-tester collaboration, FizzBuzz, prime checking, Pygame games, weather retrieval, stock analysis, reports, and itinerary planning.
- What you can do after
- Build it from scratch with AutoGen Studio, UserProxyAgent, assistant agents, group chat, or CrewAI role-based task sequences.
AI Agents in Action, Second Edition
A multi-agent system is a system in which multiple autonomous or semi-autonomous agents coordinate to perform tasks or achieve shared or related goals.
- How it is used
- Coordinates specialized agents when one agent faces limits in scope, parallel workload, or available context.
- How deep it goes
- Explains three architectures and separates decision-making from execution; treats enterprise security, governance, and scaling only briefly.
- Examples
- Uses specialized customer-support roles, concurrent research or document processing, and a research agent choosing sources to summarize.
- What you can do after
- Build it from scratch: choose and combine coordination patterns, assign decision and execution authority, and define communication between agents.
Read how AI Agents in Action, Second Edition explains multi-agent system →
Building Reliable AI Systems
A system composed of multiple AI agents that coordinate to perform tasks.
- How it is used
- Coordinates specialized agents to answer customer requests spanning product search, policy knowledge, and vision processing.
- How deep it goes
- Works through specialist agents, shared state, conditional routing, and response synthesis with a LangGraph implementation.
- Examples
- Uses a shopping assistant that combines product search, policy Q&A, and vision-based search for customer queries.
- What you can do after
- Apply it via LangGraph to route specialist agents, preserve their results, and synthesize a customer-facing answer.
Read how Building Reliable AI Systems explains multi-agent system →
The Art of AI Product Development
A multi-agent system is an AI system in which multiple agents perform specialized tasks and coordinate toward a shared goal.
- How it is used
- Coordinates specialized agents to address complex, dynamic business workflows that exceed a single agent’s capabilities.
- How deep it goes
- Explains specialization, orchestration, communication, and supervision, but stops short of implementation details.
- Examples
- Uses product discovery, prioritization, roadmap planning, customer-feedback analysis, and future-scenario generation as collaborative tasks.
- What you can do after
- Operate/configure specialized agents incrementally, monitor gaps, and add human review when confidence is low or errors could be catastrophic.
Read how The Art of AI Product Development explains multi-agent system →
Grokking AI Applications
An AI system composed of multiple agents that collaborate or coordinate to accomplish tasks.
- How it is used
- Organizes complex AI workflows through specialized agents that collaborate, delegate, or execute in sequence.
- How deep it goes
- Explains motivations, benefits, and sequential or hierarchical structures, while deferring detailed implementation to iterative design and testing.
- Examples
- Uses planner, calculator, fact-checker, billing, support, documentation, scheduling, budgeting, and compliance agents.
- What you can do after
- Designs and evaluates role-specific agent workflows with controlled delegation, communication, and iterative testing.
Read how Grokking AI Applications explains multi-agent system →
Investing for Programmers
A multi-agent pattern is an architecture in which multiple specialized agents collaborate through coordinated and integrated reasoning.
- How it is used
- Coordinates specialized agents when complex analysis benefits from parallel work, delegated reasoning, and integrated judgments.
- How deep it goes
- Explains specialization, parallelism, orchestration, and deliberation, but does not specify implementation protocols or agent code.
- Examples
- Uses an investment workflow and a courtroom-style panel of optimist, skeptic, and adjudicator agents.
- What you can do after
- Evaluate multi-agent investment-analysis designs by mapping specialized roles to coordinated recommendations.
Read how Investing for Programmers explains multi-agent system →
Build an AI Agent (From Scratch)
A multi-agent setup is an architecture in which multiple agents cooperate to complete a task.
- How it is used
- Divides complex tasks among specialists when a single agent’s prompts and tool choices become unwieldy.
- How deep it goes
- Details three collaboration patterns, their control and context behavior, and workflow variants; leaves A2A at an interface-level summary.
- Examples
- Uses technical blog writing split among research, coding, drafting, and review specialists.
- What you can do after
- Build it from scratch by composing specialized Agent instances with Workflow, AgentTool, or transfer-based orchestration.
Read how Build an AI Agent (From Scratch) explains multi-agent system →
AI Agents and Applications
A multi-agent system is a software system in which multiple agents independently or cooperatively perform tasks and pursue goals.
- How it is used
- Coordinates specialized agents to answer domain-specific requests, perform actions, and combine results across complex workflows.
- How deep it goes
- Explains router and supervisor architectures, graph assembly, communication, tracing, and reliability practices rather than agent internals.
- Examples
- Uses travel information, accommodation booking, weather, hotel availability, customer service, research, and trading workflows as examples.
- What you can do after
- Build it from scratch with LangGraph using router or supervisor patterns, specialist tools, structured state, and tracing.
Read how AI Agents and Applications explains multi-agent system →
Related concepts
- LangGraphused by · LangGraph is a framework for building stateful, graph-structured workflows and applications, including those involving language models and agents.
- AI agentuses · An AI agent is a computational system that uses artificial intelligence to interpret information, select actions, and pursue goals through interaction with an environment.
- specialized agentuses · A specialized agent is an agent focused on a distinct task, class of tasks, or domain, with responsibilities or capabilities suited to that focus.
- agent orchestrationhas part · Orchestration is the coordination and sequencing of agents, services, tasks, or components to carry out a larger workflow or objective.
- software artifacthas part · A software artifact is a persistent work product or object produced or maintained during software development or a related workflow.
These guides are generated from the text of Manning books by Manning's knowledge graph and link to the sections they draw on. The full books are available on liveBook and at manning.com.