Why I built a native control room for Claude Code agents

Hey Product Hunt 👋 I'm Jhol, and I built devpit because I kept losing track of my own agents.

Five terminals in, I no longer knew what was running, which one I was reading, or what any of it had cost. devpit is the app I wanted for that:

• One target terminal per project. Switching work switches what's attached; everything else keeps running in the background, even after you close the window (tmux underneath, psmux on Windows).

• A board where moving a card runs real work. Each column composes its own step: which agent, which model, which slice of the card's context, and a spending cap. When it finishes, the real cost is written on the card.

• An orchestrator, when you want one. A chat that sees every project you link to it, starts sessions on a card's work and tells you what they reported. Every question waiting on you lands in one inbox. It never moves a card into a lane that runs a step, and it can't approve anything on your behalf.

• The island. A small capsule above your other windows shows what each agent is doing, and lets you answer the one that's waiting without hunting for its terminal.

• Reminders that only remind. "Remind me at three to review the PR" — devpit tells you then, and starts nothing.

• Local-first. Projects, boards and history live in ~/.devpit on your machine. The account is optional.

It drives Claude Code today, runs on macOS, Linux and Windows, and it's free and open source (Apache 2.0).

I'd love to hear how you run your agents today, and what you'd want devpit to keep in view for you.