I've been kicking the tires on SuperAGI for a few weeks now, trying to set up a basic multi-agent workflow for some internal security tool integration. The promise is there, but the actual orchestration UI feels like it was designed by someone who's never actually had to *orchestrate* anything under time pressure.
Everything is a modal on top of a modal. Is the "agent" the core unit, or is it the "workflow"? Half the configuration options seem to live in a YAML file anyway, so why is the UI so busy yet uninformative? I keep asking myself:
* Why does dragging a "tool" to an agent sometimes create a link and sometimes just highlight the node?
* Where is the clear visual distinction between a *running* agent in a workflow and one that's just configured?
* The documentation says "real-time monitoring," but the logs seem to update in chunks, not streams. How am I supposed to know if an agent is stuck in a loop *right now*?
I get that it's a complex product, but complexity in the backend shouldn't mandate confusion in the UI. For a tool that sells itself on "orchestration," the control panel feels less like a conductor's podium and more like a puzzle box.
Or is this just me? Has anyone else actually built a production workflow with more than three agents and found the UI intuitive? I'm genuinely curious if I'm missing a paradigm shift here, or if the UX just hasn't caught up to the ambition of the platform.