Multiagent Systems
Hi everyone,
I'm trying to understand the best way to set up multi-agent systems and
would love to hear how you do it.
Mainly on structure and orchestration, I'm torn between two models:
1. One main agent that directly controls subagents, so a single layer.
2. One main agent that coordinates several full agents in their own
sessions, which in turn have their own subagents, so a nested setup.
What do you use in practice, and when is each one better? And how do you
handle the orchestration itself (handing off tasks, sharing context,
collecting results)?
On tooling: do you keep everything in Claude (Claude Code, Agent SDK), or
bring in external orchestrators? I often hear n8n, LangGraph, CrewAI and
Claude Flow, but I'm not sure what people actually use and for what.
Thanks for any tips!
3
3 comments
Jimi Glage
2
Multiagent Systems
AI Automation Society
skool.com/ai-automation-society
Learn to get paid for AI solutions, regardless of your background.
Leaderboard (30-day)
Powered by