Agents & Tools · advanced
Multi-Agent (multi-agent system)
A multi-agent system uses several LLM agents that talk to each other — a manager and workers, a debate, a pipeline — instead of a single agent doing everything.
Explanation
Patterns include hierarchical (one planner, several specialists), peer (agents debate or vote), and pipeline (each agent transforms the output of the last). The promise is parallelism and specialization; the cost is more LLM calls, more places for failure to compound, and harder debugging.
Empirically, multi-agent systems beat single agents on complex tasks (research, deep coding) but lose on simple ones where the overhead dominates. Real production systems often have 2-4 agents, not the 20-agent fantasies of early autonomous-agent demos.
Examples
- A planner agent + executor agents in Claude Code.
- Constitutional AI debates between two model instances.
Frequently asked
What is Multi-Agent?
A multi-agent system uses several LLM agents that talk to each other — a manager and workers, a debate, a pipeline — instead of a single agent doing everything.
What is an example of multi-agent?
A planner agent + executor agents in Claude Code.
How is Multi-Agent related to Agent?
Multi-Agent and Agent are both agents & tools concepts. An AI agent is an LLM-driven system that decides which actions to take, executes them via tools, observes the results, and iterates until a goal is met.
Is Multi-Agent considered advanced?
Multi-Agent is generally considered advanced-level material in the AI and LLM space.