I've been running into a wall with AI coding assistants. They're great individually, but I end up being the human middleware - copy-pasting between Claude windows, manually coordinating who works on what.

So I built something: Orcha (orcha.nl). It lets you run multiple Claude Code agents in parallel, each on separate git branches. Visual workflow builder to define hand-offs. All 100% local.

What took 2 hours now takes 20 minutes.

Curious: Has anyone else tried multi-agent setups? What worked? What didn't?