Multi-agent system
A multi-agent system is a setup in which several AI agents, each with its own instructions, tools and context, work on parts of a task and pass results to each other or to a coordinating agent.
Also called: multi-agent
By Divya Sharma · Checked Nov 15, 2026 · How we review
One agent with one context window can only hold so much. Splitting a task across agents lets each one focus on a piece and work in parallel.
How it works
The most common shape is orchestrator and workers. A lead agent reads the task, breaks it into parts and starts subagents to handle each part. Each subagent has its own context, does its search or coding, and returns a short result. The lead combines the results. Other shapes include a pipeline, where agents hand off in a fixed order, and peer agents that talk over a protocol such as Agent2Agent.
Example
Anthropic's research feature uses a lead agent that plans and spawns subagents to explore different parts of a question at the same time. Anthropic reports that this design used about 15 times more tokens than a chat.
Limits
Coordination is hard. Without clear task descriptions, agents duplicate work or leave gaps, and early versions of Anthropic's system started far too many subagents for simple queries. The extra tokens only pay off when the task is valuable and can be split cleanly. Many coding tasks share too much context to split well.
See agent orchestration for how the flow between agents is controlled, and CrewAI for one framework built around teams of agents.
Related terms
- Agent orchestration
- Agent orchestration is the control of which agents or model calls run in an application, in what order, and how each next step is chosen, either by a model making the decision or by fixed code.
- Agent2Agent protocol
- The Agent2Agent protocol (A2A) is an open standard that lets AI agents built on different frameworks and by different vendors find each other, delegate tasks and exchange results without exposing their internal tools or memory.
- Subagent
- A subagent is a helper agent that a main agent hands a task to. It works in its own fresh context window, often with its own instructions and tools, and returns only a summary of the result.
- AI agent
- An AI agent is a system in which a language model decides its own next steps: it calls tools, reads the results and repeats until the task is done, instead of following a fixed script.
- Agent framework
- An agent framework is a software library that provides the common parts of an AI agent, such as the tool loop, state, memory, handoffs between agents and tracing, so developers do not build them from scratch.