Qwen Code in your MacBook notch

Three agents, one place to look

Running Qwen Code usually means running it next to something else - one agent on the refactor, another on the tests, a third on whatever you fired off and forgot. Each of them has its own terminal, its own prompt style, and its own way of going quiet.

That is the case CrewTower was built for. Every running session, from every agent, is one card in one strip. The waiting ones come first. There is no per-agent window to arrange and no rotation to keep up.

Knowing which one stopped

With one agent, "something is waiting" is enough information. With four, it is not. Each card carries its agent, what that session is doing, and how long it has been going, so you know what you are answering before you answer it - and Jump takes you to the exact window, tab or split pane that session belongs to, not merely to the terminal application it happens to be in.

Answer in place

When an agent stops for a decision, its hook blocks on a local Unix socket. CrewTower answers over that same socket, so the prompt resolves where it is: no terminal comes forward, nothing is typed. There is deliberately no synthetic-keystroke path in the app, which is what makes answering four sessions in a row safe to do quickly.

Destructive shell commands are flagged on the card first. Two ways to say yes: once, or every permission request for the rest of that session, a grant CrewTower keeps itself rather than a rule written into Qwen Code's settings.

The rest of the crew

Beyond the six it speaks natively, CrewTower recognises twenty more agents, from GitHub Copilot and Factory Droid to Aider and Amp, and gives each one a card you can jump from. You lose the decision channel, not the visibility; the agents page lists them all, with what CrewTower does for each.

Terminals it lands in

Jump lands on the exact tab in Terminal, iTerm2 (down to the split) and Ghostty; the pane in WezTerm, tmux and herdr, and in kitty with its remote control on; Warp's tab through the focus link Warp provides; and the window in Alacritty, which has no tabs. A terminal with no built-in support can be taught with one line containing placeholders, and a custom rule always beats the built-in one. How Jump finds the exact tab is written up terminal by terminal. Answering from the notch never needs macOS Accessibility permission; Jump needs it only in some terminals, such as Ghostty and Alacritty.

Setup is one launch

Hooks are configured on first launch, for every supported agent at once, and cleaned up when you remove the app.

It stays on your Mac

Sessions are read from what the agents already write locally and decisions travel over a local socket. No accounts, no cloud, no analytics in the app, and no session visible from another machine.

What it does not do

CrewTower's usage meter shows how much of your Claude Code or Codex subscription window is left, and how much context each session has left, but it keeps no token or cost history by day, session or model, and shows no cost for Qwen Code sessions - ccusage is the tool for that. It is not a terminal, not a cloud dashboard, and not a way to run agents. It watches and answers the ones you started yourself.

Questions

Does CrewTower work with Qwen Code?

Yes. Qwen Code is one of the agents CrewTower speaks to natively, with full status and decision support wherever the agent exposes a decision channel.

Can I run Qwen Code alongside other agents and see them all?

Yes. Every running session gets its own card in the notch regardless of which agent it belongs to, and the ones waiting on a human sort to the front so the row stays useful when there are many. The twenty other agents CrewTower recognises get a card you can jump from too.

How do I tell which session is which?

Each card carries its agent and what that session is doing, and Jump focuses the exact terminal window, tab or split pane it belongs to rather than just the terminal application.

Does answering one session disturb the others?

No. Each answer goes back over that session's own socket connection. Nothing is focused and nothing is typed.

Do I need a separate setup per agent?

No. First launch handles the supported agents together.

Does CrewTower report Qwen Code usage or cost?

CrewTower has a usage meter, but it covers the Claude Code and Codex subscription windows rather than Qwen Code. For Qwen Code sessions it shows how much context is left. It keeps no token or cost history by day, session or model, and shows no cost for Qwen Code sessions. For that, ccusage is the tool people use.

Which other agents are spoken natively?

Claude Code, Codex, Cursor, Gemini CLI and opencode. The twenty other agents CrewTower recognises get a card you can jump from, and the agents page lists all twenty-six with what it does for each.