Omarchy Quattro pre-wires every coding-agent CLI and refuses to name a default

Omarchy's latest release, Quattro, pairs the Hyprland desktop with coding agents as a shipped part of the system, and its manual is explicit that it does not choose among them: "every major coding-agent CLI comes pre-wired as a lazy-loaded launcher" ^[1][2]. That is a more interesting design decision than the AI branding around it, because it is an early attempt to answer how an operating system hosts agents without tying the distribution's fate to one vendor.

The mechanism is deliberately unglamorous. According to the manual, the launchers are small mise-managed stubs in ~/.local/bin/, and nothing is downloaded until the first time a user actually runs one ^[2]. The cost of shipping ten vendors is therefore close to the cost of shipping one, and the choice, along with the account and the billing relationship, is deferred to the person at the keyboard rather than baked into an image.

The neutrality has a boundary, and it sits at runtime rather than at install. Quattro watches systemd-coredump, raises a "Process crashed" notification, and on click hands the crash to your default agent along with a built-in diagnose-crash skill that walks the agent through reading the core dump and deciding whether the failure is worth reporting ^[6]. Agent skills also cover system tailoring, Hyprland config edits, top bar changes, and building a theme from scratch ^[7]. So the packaging layer commits to nobody while the integration layer assumes exactly one selected agent ^[1]. That is a coherent split, but it means the interesting compatibility work is in the skills, not in the launchers.

The onboarding, in ZDNet's account, is where the seams show. Setting up Gemini opened an authentication browser window trapped behind the agent setup overlay, which the reviewer could only reach by opening a third application and forcing the windows to tile ^[3]. Google account login then failed to authenticate the agent, and an API token had to be created instead ^[4]. After that, the reviewer reports the agent worked, functioning as a command-line chat client with support for sandboxes ^[5].

Two other pieces matter to operators watching spend. An agent icon appears on the top bar the first time Omarchy detects AI coding use on the machine, and it tracks subscriptions including plan, usage percentage, weekly limits, and token usage by day and by model ^[8]. Local models are handled separately, through Install > AI > LM Studio in the desktop menu; the review notes LM Studio ships a GUI where Ollama installs only the CLI, and that LM Studio can run alongside the default CLI agent ^[9][10].

None of this is aimed at a general audience. ZDNet's reviewer is blunt that this is a developer distribution rather than one for the average user, and that Hyprland demands memorised keybindings, starting with Super+W to close a window and the Learn > Keybindings menu to find the rest ^[11][12].

Watch three things. Whether the stub-launcher pattern gets copied by other distributions, since it is cheap to implement and easy to audit. Whether the diagnose-crash and system-tailoring skills hold up across agents as vendor CLIs change shape, because that is where single-vendor assumptions creep back in. And whether the subscription tracker's plan and limit accounting stays accurate as providers rewrite their pricing.