Ask HN: What is the underlying stack behind multi-agent platforms?
Recently, I am seeing lots of startups with multi-agent platform, where you can create your own agent template, attach tools and run it reliably.
Which frameworks, platforms are you using for these kind of multi-agentic platforms?
We are using LangGraph, which is quite low level, but has useful features like time travel, human in the loop capabilities (interruptions), flexibility on the paradigm.
How do you run it reliably/durable?
Does it have automatic checkpoints to be able to resume conversations?