Multi-agent refers to a configuration in which several agents interact to accomplish a shared goal. The concept originates in distributed artificial intelligence research on multi-agent systems (MAS), where autonomous entities cooperate or compete; with the spread of agents built on large language models (LLMs), the term has come to describe designs that assign different roles to multiple LLM agents.
A typical setup places role-specific agents for planning, coding, testing, or reviewing under an orchestrator agent that divides the work and collects the results. Having one agent verify another's output to reduce hallucination or overstatement is another common pattern. Frameworks such as Microsoft's AutoGen, CrewAI, and LangGraph are widely used to build multi-agent configurations.