Running one AI coding agent at a time feels wasteful when you have three features to build. Kage is a new open-source terminal UI that lets you spin up multiple AI coding agents side by side, each in its own tmux session with an isolated Git worktree.
The setup is straightforward: each agent gets its own copy of your repo via Git worktrees (separate working directories that share the same Git history), so they can write code simultaneously without stepping on each other's changes. Tmux handles the session management, and Kage's TUI sits on top to let you monitor and switch between agents. It works with Claude Code, Codex, and similar terminal-based AI coding tools.
This solves a real friction point. Anyone running Claude Code or similar agents knows the pattern: start a task, wait, review, start the next one. Kage lets you farm out multiple tasks at once and check in on each agent's progress. The Git worktree approach is smart because it avoids the merge conflicts you'd get if two agents were editing the same working directory.
It's early-stage software and clearly aimed at developers comfortable with tmux and Git. If you're already running AI coding agents from the terminal, the workflow overhead is minimal. If tmux isn't part of your daily toolkit, the learning curve stacks up fast.