OpenAI Brings ChatGPT and Codex to Linux: The Native Desktop App Has Arrived

For years, Linux developers watched from the sidelines as macOS and Windows users enjoyed native ChatGPT desktop apps. The community posted feature requests, signed petitions, and built unofficial wrappers — all asking the same question: When is OpenAI going to support Linux? On August 11, 2026, OpenAI finally answered. The company released a ChatGPT desktop app for Linux in preview, bringing ChatGPT, ChatGPT Work, and Codex together in a single native application available worldwide.

“Linux has been one of the most-requested platforms for the desktop app, and this launch extends ChatGPT and Codex across every major desktop operating system,” an OpenAI spokesperson told The New Stack. It is a milestone that closes a long-standing gap in OpenAI’s desktop strategy — and signals that the company now treats Linux as a first-class platform rather than an afterthought.

What the App Delivers

The Linux desktop app is not simply a port of the Mac or Windows version. It is a unified workspace designed around three core pillars: the standard ChatGPT chat interface, ChatGPT Work for knowledge-work tasks, and Codex for agentic software development. Users can switch between these modes within a single window, managing multiple projects, inspecting generated artifacts, reviewing code changes, and orchestrating long-running agent tasks that may take minutes or hours to complete.

The app includes several features that power users have come to expect from the desktop experience. An in-app browser allows ChatGPT to navigate the web without leaving the application. Chrome extension support means that existing browser-based workflows and plugins continue to function. Voice input is supported natively, letting users interact hands-free. And the project dashboard view provides a visual coordination layer for juggling multiple repositories, cloud-based tasks, and local work simultaneously.

For developers who prefer the terminal, the desktop app does not replace the existing Codex CLI. The CLI remains the more direct tool for inspecting repositories, running commands, scripting repeatable operations, and integrating Codex into CI/CD pipelines. Instead, the desktop app targets the coordination layer around those tasks — the place where you need a GUI to track several agents running in parallel, approve code changes, and review multi-file diffs.

Supported Distributions and Architectures

OpenAI has been deliberate about which Linux distributions it supports at launch. The preview runs on:

• Ubuntu 24.04 LTS and Ubuntu 26.04 LTS (desktop variants)
• Debian 13
• Fedora 43 and Fedora 44

Packages are distributed as native .deb files for Debian-based systems and .rpm files for Fedora, covering both x64 and ARM64 architectures. This dual-architecture support is significant: it means developers running on ARM-based systems like the Raspberry Pi 5 or ARM workstations are not left behind. The choice of Ubuntu, Debian, and Fedora as the initial target set also ensures broad downstream compatibility, since hundreds of derivative distributions — including Linux Mint, Pop!_OS, Kubuntu, and others — share package compatibility with these upstream bases.

The decision to support Fedora 43 and 44 is particularly notable, given that some earlier community reports and sign-up forms had listed Fedora as unsupported. The final release resolved this ambiguity, bringing Fedora into the fold alongside the Debian family.

The Feature Gap: No Native Computer Use

While the Linux app delivers the core ChatGPT and Codex experience, it arrives with one significant limitation. Native Computer Use — the feature that allows ChatGPT to interact directly with desktop applications like LibreOffice, GIMP, or other GUI programs — is not available at launch. This gap has downstream effects on several related capabilities:

• Appshots (screenshots of application state) are unavailable
• Record & Replay functionality is not supported
• Voice requests involving desktop applications do not function
• Browser-based workflows, however, remain fully functional through the in-app browser and Chrome extension

An OpenAI spokesperson confirmed: “The Linux app supports the core ChatGPT and Codex experience, including Voice. The main difference is that native Computer Use — the ability for ChatGPT to interact with desktop applications like OpenOffice or GIMP — isn’t available at launch. This also affects related features, including Appshots, Record & Replay, and voice requests involving desktop apps and browser.”

This limitation is unlikely to be permanent. The macOS and Windows versions of the app include full Computer Use support, and the underlying technical challenge on Linux — navigating the fragmentation of display servers (X11 versus Wayland), window managers, and accessibility frameworks — is well understood. But for now, Linux users who rely on the app will need to work within this constraint.

Context: The Broader Desktop AI Race

The Linux launch does not happen in isolation. OpenAI’s desktop app strategy has accelerated rapidly through 2026. The Mac and Windows versions launched in July alongside the introduction of ChatGPT Work, OpenAI’s agentic knowledge-work tool. At that time, the desktop app felt like a rebranded Codex client with ChatGPT chat tacked on. Since then, OpenAI has refined the interface, making it easier to move between chat, work, and coding modes.

More importantly, OpenAI is now competing directly with Anthropic on the Linux desktop. Anthropic released its Claude desktop app for Linux roughly a month before OpenAI’s launch, supporting Ubuntu 22.04+ and Debian 12+. Claude’s app includes its own agentic coding capabilities through Claude Cowork, and the competitive dynamic between the two companies is now playing out across all three major desktop platforms.

The broader industry trend is clear: AI companies are racing to own the developer’s desktop. The terminal was the first battleground (CLI tools), followed by the IDE (extensions), then the cloud (managed agent platforms), and now the native desktop application layer. Each surface offers different affordances for agent coordination, and companies are betting that the desktop app — with its visual workspace, file management, and multi-project orchestration — will become the primary interface for working with AI agents.

Why This Matters for the Open-Source Community

The Linux developer community has historically been skeptical of proprietary AI tools. Many Linux users prefer open-source alternatives, self-hosted models, and tools that respect the Unix philosophy of small, composable programs. OpenAI’s decision to invest in a native Linux app acknowledges that this community remains a critical audience — not just for adoption metrics, but for the developer feedback loops that shape product direction.

Early community reception has been mixed but generally positive. Reddit discussions highlighted technical concerns about the app’s underlying framework (Tauri, which uses webkit2gtk on Linux and has known stability issues), while the OpenAI Developer Community forums saw enthusiastic adoption among developers eager to try Codex in a native context. The GitHub repository for an unofficial wrapper, ilysenko/codex-desktop-linux, saw a spike in attention before users migrated to the official app.

Looking Ahead

The preview label means features will evolve. OpenAI has not announced a timeline for full release or for closing the Computer Use gap, but the company’s track record suggests rapid iteration. For now, the Linux desktop app represents a significant step: the world’s most prominent AI company has finally given the Linux community what it asked for — a seat at the table, with all the capabilities and caveats that entails.

As agent-based workflows become the default for software development, the desktop application layer will only grow in importance. OpenAI’s Linux launch is not just about catching up to Anthropic. It is about ensuring that when developers choose their primary AI workspace, ChatGPT and Codex are available wherever that developer happens to be — including the penguin.