HomeGlossaryMulti-agent team

Multi-agent team

Also called: agent team, multi-agent system

Definition

A multi-agent team is a group of agents that run together and communicate, each handling part of a task rather than one agent handling all of it.

In practice

The argument for splitting work across agents is the same as for splitting it across people: narrower instructions produce more reliable behaviour than one long brief covering everything. A researcher, a writer, and a reviewer each do one job well. The argument against is that every additional agent multiplies cost, latency, and the number of places a failure can hide. Start with one agent and split only when a specific step is demonstrably being done badly. A team assembled before there is evidence of a problem is usually a slower, more expensive version of a single agent.

Example

A content agency splits its weekly newsletter across three agents in one shared workspace. A researcher gathers the week's sources, a writer drafts from the researcher's notes, and a reviewer checks every claim against the sources before anything reaches an editor. The split came after a single agent kept inventing citations: giving the checking step its own narrow job fixed it, and nothing else was split.

Common questions about multi-agent team

When should I use a multi-agent team instead of one agent?

Use a multi-agent team when a specific step is demonstrably being done badly and a narrower instruction would fix it, such as separating fact checking from writing. Start with one agent. Each extra agent adds cost, latency, and places for failures to hide, so split only when there is evidence.

How do agents in a multi-agent team work together?

Agents in a multi-agent team each handle part of a task and pass work between them, often through a shared workspace or messages. A common pattern is a pipeline, such as researcher, writer, and reviewer, where each agent has one narrow job and the next one checks or builds on its output.

Read more

Nearby terms

All 25 terms

Run agents without the homework

Qoren runs OpenClaw, Hermes, and Codex agents in managed environments, with the restarts, secrets, schedules, and spend caps handled for you.