How we build Beadbox using Beadbox

We run 10+ Claude Code agents in parallel to build this product (engineer agents, QA, product, marketing, a release agent). All coordinated through the beads issue tracker, all visible in Beadbox in real time.

It's a strange feedback loop. Every time the epic tree view is confusing, we feel it first. Every time dependency badges are hard to read at 50+ issues, we hit that wall before anyone else does.

Happy to share how the multi-agent setup works if anyone's curious, it's a pretty different way to run a software project.

What are you using to track work when you have AI agents filing issues and pushing code?