Can Thurbox run Doom?
Yes.
Now that the important question is out of the way: Thurbox exists to orchestrate coding agents and give you an efficient way to work with them. On top of that, it fills in features some agents are missing — and it does so in an agent-agnostic way. Sending and receiving messages between agents, for instance.
- Website: https://thurbox.thurbeen.eu/
- Repo: https://github.com/Thurbeen/thurbox (Rust, MIT)
Why a CLI and a TUI instead of an app
Most of us worked this out a while ago: a plain UI application isn't enough once you're orchestrating multiple tasks every day. What developers and power users actually want is a CLI that can set up and automate agentic sessions.
But once those sessions exist, you still need somewhere to watch them and step in. Hence the Thurbox TUI.
A terminal interface doesn't have to mean keyboard-only, either. Thurbox supports the mouse — click a session to select it, click into a pane, scroll — and more generally it tries to do as much as it can to stay usable for everyone, whatever way you prefer to drive it.
Why I built it
I'm a DevOps engineer, and I work across a lot of projects at once — many repos, many contexts.
I built Thurbox to cut the mental overhead of managing several agentic sessions at the same time: remembering to claude --resume in the right repo, juggling worktrees, keeping track of what's running where.
How I work with it today
My workflow keeps evolving toward multi-agent orchestration. Right now it looks like this:
- One fleet agent handles all my tasks from a single session.
- It communicates through Thurbox tasks and messages with a dispatch agent.
- The dispatcher spawns new sessions with the right context to resolve each issue.
I pair that with a custom Thurbox layout and panes, so I can monitor progress on each topic at a glance.
The part I find most satisfying
Closing the UI doesn't kill anything. Sessions are backed by tmux, so I can quit Thurbox, go do something else, and every agent keeps working in the background.
Same story across a reboot. I restart my machine, run thurbox, and I'm back where I left off — the session list, the repos and worktrees they're attached to, and my customizations too. The panes, the layout, the theme: they're plain files in a directory I own, so nothing to set up again.
It also means Thurbox itself gets out of the way. I can upgrade it — or roll back to an older version — while my agents keep running underneath. And I can keep tinkering with my custom UI, moving panes around and rewriting layouts, without interrupting a single session.
That all sounds like a small thing. In practice it's the difference between a tool you open when you need it and one you just leave running.
What Thurbox is not
Thurbox is not another tmux reimplementation, the way something like herdr is. It's also not a heavy web app running in the background, like VS Code agents, Cursor or Orca.
Thurbox runs on battle-tested software as its core dependencies. Which means that if Thurbox breaks, your sessions are still there — and you can go back to raw tmux to handle them.
Install it
You'll need tmux 3.2+, git, and at least one coding-agent CLI (Claude Code, Codex, opencode, aider, Copilot CLI, and others are supported out of the box).
On Linux and macOS:
curl -fsSL https://raw.githubusercontent.com/Thurbeen/thurbox/main/scripts/install.sh | sh
On Windows:
irm https://raw.githubusercontent.com/Thurbeen/thurbox/main/scripts/install.ps1 | iex
If you'd rather use a package manager: brew install thurbeen/thurbox/thurbox, winget install Thurbeen.thurbox, choco install thurbox, or paru -S thurbox-bin on Arch.
Then just run it:
thurbox
That's the whole setup — sessions, agents, themes and the interface itself are seeded on first launch. From there, Ctrl+N creates a session: pick your repos, name it, choose an agent, and optionally spin it up on a fresh git worktree.
