◆ An open-source experiment
That is no longer how software gets built.
Coding agents can work across multiple tasks, repositories, terminals, and review cycles at once. But today's tools still make you manage that work through scattered windows, tabs, chats, and terminal sessions.
Rook is an experiment in what comes next: an agent-native IDE built around persistent workspaces, concurrent agent sessions, review, and attention management.
It is early. The shape of this new category is not settled yet.
One developer. One repository. One task at a time. Every tool we use — the editor, the terminal, the review UI — quietly assumes it. That assumption is breaking.
When several agents work in parallel across repositories, the bottleneck stops being how fast you can type and becomes how well you can direct, review, and stay oriented across all of it at once. That is the problem rook is built around.
Rook is a real terminal you can install now — a desktop app that replaces ghostty + tmux, built for muscle-memory parity first. The agent is being built from inside a tool that is already trusted. Parity first, magic second.
Spawn an agent, watch it work, and stay oriented across the whole session — from one workspace.
If the tools you have today do not fit the way you actually work with agents, come help define what replaces them.