编排差距：为什么你的 AI 代理无法相互发现

Building multiple AI agents can lead to challenges when trying to effectively chain them together. Each agent may perform well in isolation, but the core issue arises from the fact that these agents are unaware of each other's existence. When a call to another agent is hardcoded into the code of one agent, it limits the flexibility of the multi-agent system by establishing a fixed relationship at design time. This hardcoding complicates the process of replacing or adding agents, impacting the system's scalability and adaptability. Addressing this issue requires a rethinking of how agents interact to achieve true collaboration and dynamic responsiveness.