Building a CLI agent for Claude to manage How to design a multi-agent system where Claude delegates work?
一个开发者社区正在讨论如何设计一个多智能体系统，其中大型语言模型（LLM），特别是Claude，可以管理另一个工作智能体。核心问题是确定最适合这种委托的框架或架构方法，例如MCP、自定义CLI封装器或专门的API循环。社区正在寻求关于此主题的看法、设置或建议。
- 发布
- 09/07 14:49 UTC+0
- 收录
- 09/07 19:00 UTC+0
- 来源类型
- 开发者社区
- 档位
- 社区
- 信源状态
- 正常
档位是按信源手工设定的编辑判断，不是逐条打分。
Hey everyone,
I’ve been experimenting with a terminal-based workflow where I use Codex to execute tasks and Claude to manage the overall project. Right now, I just write my project prompts or requirements directly into the terminal, and they handle everything.
However, I feel like they struggle a bit with high-level architectural design and UX flow.
I want to build my own custom AI agent that works similarly specifically, an agent that Claude itself can call, manage, and delegate tasks to via the command line.
Has anyone built a setup where Claude orchestrates a secondary sub-agent in the terminal?
What frameworks or architectural approaches (like MCP, custom CLI wrappers, or specialized API loops) are best suited for letting an LLM manage another worker agent?
Any tips on improving the design capabilities of these agents when operating in a terminal environment?
Would love to hear your thoughts, setups, or recommendations!