Running multiple AI coding agents shouldn't feel like herding cats

I built Agent World because I was running several Claude Code agents at once and kept losing track of them across terminal tabs: which one was waiting for me, which one was about to edit the same file as another.

What it does:

• Each agent gets its own git worktree and branch, so agents can work on one repo in parallel without overwriting your files or each other's.

• Plans, permissions and questions show up in one place. Approve a plan or allow a command in one click, without hunting for the right terminal.

• Workflows (optional): plan → build → review → pull request, and the pull request waits until you press Run.

• A shared library and playbook: skills and instructions kept once and given to Claude Code and Antigravity agents.

• Usage: limits, spend, tokens and disk, by account and project.

• And yes, it's a pixel-art office where each character shows what its agent is really doing. If you'd rather not see it, ⌘⌥O hides it.

It runs your own claude, unmodified, and changes nothing global in ~.claude.

Honest limits: it's Mac only for now, and the app isn't signed with an Apple Developer ID yet. Homebrew is the easiest install:

brew tap hodkep/tap && brew trust --cask hodkep/tap/agent-world && brew install --cask hodkep/tap/agent-world

It's open source (Apache 2.0): github.com/hodkep/agent-world

I'd love to hear how you run several agents today, and what would make you switch. I'll be here all day answering questions 🙏