I think it really comes down to the complexity of the product. One agent sounds simpler for the user, but specialized agents could handle specific tasks much better. I’d probably start with one strong agent and only split things up when there’s a clear reason to. No point building a whole team of agents before you actually need one.