Built a Claude-Native terminal as a fun side project. It's called Yardmaster. We have been running three, four, five Claude Code sessions at once — but our terminals had no idea. To iTerm, a Claude that's been grinding for 20 minutes, a Claude waiting for permission to run migrations, and a dead pane all look identical: a rectangle of text. Moving between these tabs is another problem and lot of times you would even wonder which project is this instance of claude handling. And yet nobody wants to leave the terminal. Devs live here. The terminal doesn't need replacing — it needs to learn what an agent is. So Yardmaster is a real macOS terminal where Claude sessions are first-class: → Every session named, grouped by project, with live status — working, waiting on you, idle — straight from Claude Code's hook protocol. No output scraping, no guessing. → One-keystroke git worktrees: run several Claudes on the same repo without collisions, then diff, merge, or open a PR from the app. → An inbox of every session that's blocked on you, with the exact question it's asking. → Token spend per session, live on every row. → And when you just need a terminal? It's still just a fast, multi-tab terminal. Now the part that actually keeps me up at night: this took two days to build. Two days, working alongside Claude, to solve our own workflow problem — a real app, with git integration, native notifications, and a packaged installer. That's the bigger story. Software has always been take-it-or-leave-it: you use the interface the company shipped. But when an app costs two days instead of two quarters, that flips. Maybe the future isn't downloading tools and adapting to them — it's describing the tool you actually need and having AI build it around your workflow. Yardmaster is just what that future looked like for ours. Built with Claude, for people who build with Claude. Open source, free: 13 hours ago