OpenAI releases preview desktop ChatGPT app for Linux with built-in Codex

After a long wait, OpenAI releases a preview desktop ChatGPT app for Linux. The new package bundles Codex, the regular chat and Work mode under one roof, exactly what developers building on the platform expected. Users can open local repositories, run multiple code agents in parallel, review diffs and continue working in the same IDE where the code is already running.
The flow is simple: open the app, sign in with a ChatGPT account and select Codex. From there users can work directly with local files, terminals and development tools. The system lets several coding tasks run concurrently, each task keeping its own context. Git worktrees let agents work in the same repo without overwriting each other, and changes can be reviewed before proceeding. Users can ask to build features, fix bugs, refactor, run tests or explore an unfamiliar codebase. Skills allow team instructions, preferred workflows and project-specific context to be supplied. There is also an option to schedule recurring work and return when results are ready, and a built-in browser with a Chrome extension for web-based tasks. One click switches to chat for quick questions or to Work for longer tasks such as research, documentation and analysis.
The preview supports Ubuntu 24.04 LTS and 26.04 LTS, Debian 13, and Fedora 43 and 44, both x64 and ARM64. Ubuntu and Debian receive .deb packages, Fedora receives .rpm. To check architecture, x86_64 indicates the x64 package, aarch64 or arm64 indicates the ARM64 package. Installation is performed in the terminal with dpkg or rpm as appropriate, after which the application appears in the applications menu or is available via the `chatgpt` command.
The installer configures a signed OpenAI repository, so future updates arrive through the normal package manager—`apt update && apt upgrade` on Ubuntu and Debian, `dnf update` on Fedora. Wayland support remains experimental: in a Wayland session the app uses XWayland when available, and native Wayland can be forced from the terminal after a full shutdown. Features such as floating windows, window placement, focus and keyboard shortcuts may not work smoothly until support matures.
The real strength of the desktop experience is the ability to delegate multiple work pieces without losing control. One agent can investigate a failing test, a second can implement a feature, a third can review an existing change; each task stays organized and worktrees keep parallel changes isolated. Results can be inspected, corrections requested, or code opened in an editor for deeper edits. The built-in browser and Chrome extension also work on Linux, keeping web tasks available alongside local coding—whether checking documentation, examining a web app or gathering information needed for a change. Skills let these workflows become reproducible and team-specific.