多 Agent 协作Multi-Agent

Multi-Agent

解决什么问题What problem it solves

把一件复杂的事拆给多个各有所长的角色来分工完成。Split a complex job across several roles, each strong at something different.

这是什么What it is

让多个各自带不同工具和提示的 Agent 分工协作。典型模式是「规划者 + 执行者 + 审查者」。

代价是 token 消耗和延迟都会成倍上升。

Multiple agents with different tools and prompts working together — the classic pattern is planner plus executor plus reviewer.

The cost is that token consumption and latency both multiply.

为什么需要知道Why it matters

不是所有事都该多 Agent。任务简单时，一个 Agent 加工具更快也更可靠。Not everything needs multiple agents. On simple tasks one agent with tools is faster and more reliable.

现在能做什么What you can do now

出处Sources