Multi-Agent Architecture

OSS Product Help LangGraph
1

I saw the new deep agents open source library that came out and I am curious on what is the right approach for the best multi agent architecture. In many ways they seem pretty similar even though one emphasizes hierarchy but it seems like deep agents accomplishes the same thing. I know its based on use case and all that but can anyone give their opinion on what is a better standard or library to follow if we are diving deep into building and customizing agent systems using langgraph.

I have built my own react library and supervisor (multi agent library) mainly for controlling more things, getting inspiration from the supervisor repo, and I am trying to keep it up to date with the lastest developments.

2

Not sure if you figured this one out yet, but essentially when you use createSupervisor, you’ll add a key called agents, with an array of agents the supervisor will call. Use the name value you used when you used createDeepAgent, and ensure your prompt tells the supervisor when to call that agent. Good luck!