Codex gains Windows computer use and cross-device continuity

OpenAI extended Codex's Computer Use capability to Windows, allowing the agent to directly interact with Windows applications for testing, debugging and refinement inside the Codex app. The company paired this with cross-device continuity: users can kick off, review and steer agentic tasks from the ChatGPT mobile app on iOS or Android while execution continues on their Windows PC or Codex on Mac. OpenAI's official account framed it bluntly: 'Computer use now works on Windows, so Codex can take action on your Windows computer.'

The release also bundled infrastructure improvements — faster in-app browser, better stability and web compatibility — plus Codex Profiles, which surface identity, activity, usage stats and token activity so developers can monitor their own consumption. That token-tracking addition lands squarely amid the week's cost-fatigue theme.

Mechanically, Windows computer use brings Codex closer to a true autonomous desktop agent on the OS where most enterprise software lives, a meaningful expansion beyond the Mac-first agentic tooling that has dominated. The mobile-to-desktop handoff mirrors the broader assistant-to-agent pivot also visible in Google's Gemini Spark and Microsoft's async Copilots.

Competitively, this directly challenges Anthropic's Claude Code and computer-use features and Microsoft's own Copilot agent push. The differentiator OpenAI is betting on is ubiquity across ChatGPT's enormous mobile base. Watch eligibility rollout (gated to certain users at launch), enterprise security posture for an agent taking arbitrary Windows actions, and whether the in-app browser speedups hold up under real agentic workloads.