multi-agent system
In plain wordsA setup in which several AI agents, each with its own role or point of view, work together — or compete — on a problem, passing messages and results among themselves.
- 1
Language-model agents
A system made of several LLM-based agentsAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done. that interact to accomplish a task, each typically defined by a role prompt, a tool set and its own context. Common topologies are hierarchical (orchestrator and workersorchestrator–worker pattern, n.: A way of organizing AI teamwork, like a newsroom: an editor AI decides which stories a big assignment needs, sends reporter AIs to cover them at the same time, and turns what they file into one report.), sequential pipelines with handoffshandoff, n.: The moment one AI agent passes a conversation or task to another, better-suited agent — like a call-center operator transferring you to the right department., peer debate or critique, and role-played 'teams' of planner, coder, reviewer and tester. Benefits include parallel exploration, separation of concerns and cross-checking of one agent's work by another; costs include token use often an order of magnitude above a single agent, coordination failures, error amplification and harder debugging. Studies in 2025 found the gains over single agents often minimal on popular benchmarks and strongly task-dependent: with tools, prompts and compute held equal, one study measured changes from +81% on decomposable financial reasoning to −70% on sequential planning. Practitioners therefore tend to reserve multi-agent designs for parallelizable tasks.
Our multi-agent system has a planner, three coders and a reviewer, but a single agent with the same budget scored almost as well.
- 2
Classical AI
The research field, grown out of distributed AI in the 1980s and 1990s, that studies interacting autonomous agents — their communication languages, negotiation, coordination, auctions and game-theoretic behavior — in software, robotics and economics. multi-agent reinforcement learningmulti-agent reinforcement learning, n.: Reinforcement learning with several AI agents learning at once in a shared world, where each must adapt to the others — as teammates, rivals or both. is its learning-based branch.
Origin
Distributed AI took shape around 1980 (Reid G. Smith's Contract Net Protocol); multi-agent systems had dedicated conferences from the mid-1990s, merged into AAMAS in 2002. The language-model revival began in 2023 with generative agentsgenerative agent, n.: A character in a computer simulation, run by a language model, that keeps a diary-like memory, draws lessons from it and schedules its days, so that a group of them can act like a plausible small community., CAMEL, MetaGPT, ChatDev and Microsoft's AutoGen, and Andrew Ng listed multi-agent collaboration among four agentic design patterns in March 2024.
Usage note
Anthropic's June 2025 research-system post put multi-agent token use at about 15 times that of a chat. Not to be confused with a mixture of expertsmixture of experts, n.: A way to build very large AI models cheaply: the model contains many specialist sub-networks, and for each word only a few are switched on, picked by a small dispatcher network., which routes tokens among subnetworks inside a single model.
Related
- orchestrator–worker patternorchestrator–worker pattern, n.: A way of organizing AI teamwork, like a newsroom: an editor AI decides which stories a big assignment needs, sends reporter AIs to cover them at the same time, and turns what they file into one report.
- subagentsubagent, n.: A helper AI agent that a main agent starts to handle one piece of a job; it works in its own separate workspace and reports back a short summary when it is done.
- agent swarmagent swarm, n.: A large group of AI agents working at the same time on parts of a problem, usually with little central control — a name loosely borrowed from swarms of insects.
- handoffhandoff, n.: The moment one AI agent passes a conversation or task to another, better-suited agent — like a call-center operator transferring you to the right department.
- Agent2Agent protocolAgent2Agent protocol, prop. n.: An open standard that lets AI agents built by different companies find each other, say what they can do and work together on tasks, much as email lets different mail systems exchange messages.
- multi-agent reinforcement learningmulti-agent reinforcement learning, n.: Reinforcement learning with several AI agents learning at once in a shared world, where each must adapt to the others — as teammates, rivals or both.
- generative agentgenerative agent, n.: A character in a computer simulation, run by a language model, that keeps a diary-like memory, draws lessons from it and schedules its days, so that a group of them can act like a plausible small community.
- mixture of expertsmixture of experts, n.: A way to build very large AI models cheaply: the model contains many specialist sub-networks, and for each word only a few are switched on, picked by a small dispatcher network.
Sources
- Wu, Q. et al. (2023). AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation. arXiv:2308.08155. arxiv.org/abs/2308.08155
- Cemri, M., Pan, M. Z., Yang, S. et al. (2025). Why Do Multi-Agent LLM Systems Fail? arXiv:2503.13657. arxiv.org/abs/2503.13657
- Kim, Y. et al. (2025). Towards a Science of Scaling Agent Systems. arXiv:2512.08296. arxiv.org/abs/2512.08296v1
Mentioned in
AI agentAI agent, n.: An AI system that does not just answer but acts: given a goal, it decides on steps, uses tools such as a browser or code editor, checks the results and keeps going until the job is done., agentagent, n.: Something that acts. In AI, a program that observes its situation and takes actions toward a goal: a game-playing learner, a software helper, or a chatbot that uses tools to complete tasks., agent frameworkagent framework, n.: A software library that gives developers ready-made parts for building AI agents — model connections, tool handling, memory, multi-agent coordination and logging — so they need not write everything from scratch., Agent2Agent protocolAgent2Agent protocol, prop. n.: An open standard that lets AI agents built by different companies find each other, say what they can do and work together on tasks, much as email lets different mail systems exchange messages., intelligent agentintelligent agent, n.: In AI theory, a decision-maker, such as a robot, a self-driving car, a chess program or a helpful piece of software, that notices what is going on around it and responds in ways that serve its goals., multi-agent reinforcement learningmulti-agent reinforcement learning, n.: Reinforcement learning with several AI agents learning at once in a shared world, where each must adapt to the others — as teammates, rivals or both., orchestrationorchestration, n.: Coordinating several AI calls, tools or agents so each does its part in the right order and their results are combined into one finished piece of work, as a conductor directs an orchestra., orchestrator–worker patternorchestrator–worker pattern, n.: A way of organizing AI teamwork, like a newsroom: an editor AI decides which stories a big assignment needs, sends reporter AIs to cover them at the same time, and turns what they file into one report., subagentsubagent, n.: A helper AI agent that a main agent starts to handle one piece of a job; it works in its own separate workspace and reports back a short summary when it is done., The Society of MindThe Society of Mind, prop. n.: AI pioneer Marvin Minsky's 1986 book arguing that intelligence comes from huge numbers of tiny, simple processes working together, no one of which is smart by itself., agent swarmagent swarm, n.: A large group of AI agents working at the same time on parts of a problem, usually with little central control — a name loosely borrowed from swarms of insects., AI co-scientistAI co-scientist, n.: An AI system built to work alongside human scientists, proposing and debating new research ideas and hypotheses that people then test in the lab., AutoGenAutoGen, prop. n.: A free Microsoft software framework for building teams of AI agents that solve problems by messaging one another, writing and running code, and asking people for help when needed., belief–desire–intention modelbelief–desire–intention model, n.: A way of designing software agents around three human-like ideas: what the agent believes about the world, what it wants, and which plans it has committed to carrying out., blackboard systemblackboard system, n.: A design for AI programs that lets many independent specialist modules collaborate only through a shared workspace: each watches it, posts a piece of the answer when it sees a chance, and builds on what others posted., CiceroCicero, prop. n.: An AI from Meta that in 2022 played the strategy board game Diplomacy against people online, negotiating and forming alliances in ordinary written language, and scored better than most of its human opponents., CrewAICrewAI, prop. n.: A free software framework for setting up teams of AI agents, each given a job title, goal and backstory, that divide up tasks and work through them together like a small staff., generative agentgenerative agent, n.: A character in a computer simulation, run by a language model, that keeps a diary-like memory, draws lessons from it and schedules its days, so that a group of them can act like a plausible small community., handoffhandoff, n.: The moment one AI agent passes a conversation or task to another, better-suited agent — like a call-center operator transferring you to the right department., LangGraphLangGraph, prop. n.: A free software library for building AI agents as flowcharts of steps that share a memory, so developers control exactly how an agent moves between steps, pauses and resumes., MoltbookMoltbook, prop. n.: A social network opened in early 2026 for AI agents only: the bots post, comment and vote in forums much like Reddit's, and people can mostly just read along., swarm scalingswarm scaling, n.: Getting better AI results by putting many AI agents on a task at once instead of letting one think longer — faster, but more expensive for each gain.