OpenAI released Codex app version 26.527 on May 29, adding two capabilities for Windows: remote control and Computer Use. The update lets users start and steer coding sessions on a Windows PC from an iPhone, iPad, Android device or another Mac, while Computer Use enables the AI to operate desktop applications by seeing the screen, clicking and typing.
The release closes a gap left on May 14, when OpenAI first enabled mobile supervision of Codex sessions—but only for macOS hosts. By extending that workflow to Windows, OpenAI broadens Codex’s addressable user base and moves the tool beyond coding toward general-purpose desktop assistance.
Remote control turns the phone into a supervision console. Users can start new threads or continue existing work on a Windows host, send follow-up prompts, approve commands, and review outputs such as diffs, terminal output, screenshots and test results. Notifications and the ability to switch between hosts and threads keep the user in the loop without a keyboard. All execution, files, credentials and tool configurations remain on the host; the phone never runs Codex directly.
Computer Use on Windows lets Codex interact with foreground desktop apps. The AI sees the screen, clicks buttons and types text. Unlike the macOS version, which OpenAI’s earlier materials described as capable of background or locked-machine work, Codex on Windows requires the host to remain unlocked and in the foreground. This constraint may affect unattended automation workflows, though direct comparisons have not been independently verified.
OpenAI recommends Windows 11 for the full experience. Windows 10 version 1809 and newer works on a best-effort basis, provided the system is fully updated. Older builds are not recommended, and the company has not clarified whether Windows 10 support is considered production-ready.
The update caps a rapid iteration cycle. Codex began as a macOS-only coding agent in February. A Windows app arrived on March 4. On April 16, OpenAI recast Codex as a computer-use agent. The May 14 mobile launch brought remote supervision for macOS and remote environments, with Windows listed as "coming soon." The May 29 release delivers that promise and adds Computer Use.
The move expands Codex’s addressable surface and brings it closer to desktop assistant territory, a field where operating system makers are embedding their own AI features. The gap between Windows foreground and macOS background modes could shape how IT teams deploy each version—foreground-only operation may be less attractive for unattended automation tasks.
The exact rollout scope remains unconfirmed. OpenAI has not detailed whether remote control and Computer Use are available across all subscription tiers, in every region, or under enterprise administration policies. Independent verification of staged deployment was not available at press time. User testing will be needed to gauge how the foreground-only Windows implementation performs in daily workloads compared with macOS.
For technology leaders, the update signals that OpenAI is systematically closing platform gaps, making Codex a more credible cross-OS automation layer. A Windows 11 PC can now serve as a managed execution node for code and desktop tasks, supervised from a phone.