A single agent operates alone, handling all reasoning and actions. A multi-agent system combines specialists — one for research, one for analysis, another for writing or verification. These agents communicate, negotiate, and verify each other’s work, often leading to better reliability and explainability.