A multi-agent team is a setup where several AI agents, each with its own role, split up a job and pass work between them.
In plain words
Instead of one agent doing everything, you define agents with a role, a goal and some background, such as researcher, writer and editor. A lead agent or a fixed flow hands out tasks, workers may run in parallel with their own context, and results are combined at the end. Frameworks such as CrewAI and AutoGen help build this. The trade-off is cost: Anthropic reports that multi-agent systems use about 15 times more tokens than a chat.
A quick example
Imagine a news-summary crew. A researcher agent gathers articles on a topic, an analyst agent pulls out the key facts, and a writer agent turns them into a short brief. Each agent is set up with its own role and goal.
Where you'll meet it
- Research and report writing.
- Complex tasks that split into parallel parts.
Related
Sources
- DocscrewAI README, CrewAI on GitHub · read 6 Nov 2026
- DocsAutoGen, Microsoft · read 29 Sept 2026
- DocsHow we built our multi-agent research system, Anthropic · read 29 Sept 2026
Checked by WhatIsAI against the sources above · 6 Nov 2026. Spotted an error? Report it.