One of my senior was asking me try out BMAD for some time now.
So decided to try BMAD a few days back with Claude Code on a script which i had, which automates downloading files.
It was actually pretty good.
The party mode was awesome. Watching the various personas discuss among themselves felt natural. Instead of just prompting an AI and vibing through the implementation, it felt more like sitting with a small team and architecting a solution together.
It identified a few useful features and improvements that I had not thought about and then implemented them.
The BMAD helper skill was super useful too. Whenever i didnt know what to do next, it was there to guide you through.
What stood out to me was that this workflow changes the role of AI from “write code for me” to “help me think through the system”.
It felt less like autocomplete and more like collaborating with a team.
But it was definitely verbose and very documentation-oriented. Lots of planning, discussions, reasoning, structured thinking, and process. For larger or long-running projects I can see why that is useful. Especially for architecture-heavy work where context matters.
But for smaller scripts, sometimes you just want “dude just edit the file”.
Still, I can now understand why people are excited about these multi-agent / team-style AI coding workflows.
Feels like each developer is now going to get their own virtual engineering team.