Multi-Agent System
A multi-agent system is several narrower agents coordinating on one outcome, rather than a single agent holding every tool and every instruction.
The argument for splitting is control rather than capability. An agent with six tools and a tight brief is easier to constrain, test and reason about than one with sixty and a broad one. It also fails more legibly: when something goes wrong you know which agent did it and what it was permitted to touch.
The cost is coordination. Work has to be decomposed, results handed between agents without losing context, and failures handled somewhere that can see the whole run. That coordination layer is orchestration, and it is usually where the real difficulty of a multi-agent system lives — not in the agents themselves.