One month after arriving on macOS, OpenAI has released the Codex app for Windows, aiming to make agent driven software building feel normal inside real Windows workflows.
Quick Summary – TLDR:
- Codex is now available on Windows, bringing its multi agent workflow to a much larger developer base.
- OpenAI says the Mac app passed one million downloads in its first week, and Codex now has about 1.6 million weekly active users.
- The Windows version focuses on native sandboxing, letting agents run directly in tools like PowerShell, with an option to use WSL.
- Skills and Automations expand Codex beyond code writing into design to code, deployment, and recurring dev ops style tasks.
What Happened?
OpenAI has brought its Codex desktop app to Windows, following the macOS launch in early February 2026. The company says demand is strong, with more than 500,000 developers on the Windows wait list and Codex usage continuing to climb.
The Codex app is now on Windows.
— OpenAI Developers (@OpenAIDevs) March 4, 2026
Get the full Codex app experience on Windows with a native agent sandbox and support for Windows developer environments in PowerShell.https://t.co/Vw0pezFctG pic.twitter.com/gclqeLnFjr
Why the Windows release matters?
For many teams, Windows is still the default environment for enterprise development, internal tools, and a lot of production workflows. OpenAI is clearly trying to meet developers where they already work, rather than forcing a switch to a different setup.
An OpenAI spokesperson told The New Stack the Windows version was built “for real Windows developer environments,” with a focus on native workflows and familiar tooling. That positioning matters because Codex is not just an add on chat box. It is intended to be a system you run alongside real projects, real repos, and real release schedules.
A command center built around multiple agents
OpenAI describes the Codex app as “a new form factor designed as a command center for agents,” and that is the core idea behind the product. Instead of one long conversation where context gets messy, Codex lets developers run parallel agent threads across projects and tasks.
A big piece of that is isolation. Codex supports Git worktrees, so each agent can work on its own copy of the codebase. That reduces the usual friction of juggling branches and worrying about conflicts when multiple changes are happening at once.
OpenAI has also highlighted the scale of tasks it wants Codex to handle. In one demo, the company showed Codex producing a full 3D racing game with multiple maps and characters, consuming more than 7 million tokens from a single prompt.
Skills and Automations push Codex beyond code writing
Codex is built to connect to the real tools developers already use. Skills bundle instructions, scripts, and resources so agents can reliably complete repeatable workflows.
Based on OpenAI’s documentation and launch materials, Skills include integrations that help move work across the pipeline, including design and delivery steps. In practice, that means Codex is trying to be helpful not only during coding, but also during the steps that slow teams down, like turning designs into implementation tasks or packaging a release.
Then there are Automations, which target recurring chores. Examples mentioned across coverage include daily issue triage, summaries of failing builds, and release briefs that can be queued up for a developer to review.
Windows native sandboxing and open source code
Security and control are a major theme in the Windows rollout. OpenAI built a sandbox that operates at the OS level, using restricted tokens, file system controls, and dedicated sandbox user accounts. The goal is to let agents run directly in Windows environments like PowerShell without forcing people into virtual machines or WSL, although WSL is still an option for those who prefer it.
OpenAI has also published the Windows sandbox code as open source on GitHub, which may help more security minded teams evaluate how it works.
Pricing and availability
Codex for Windows is available across all ChatGPT plans, according to multiple reports and OpenAI documentation. That broad availability is part of why Codex adoption has moved quickly. OpenAI says the macOS app passed one million downloads in its first week, and overall weekly active users are now around 1.6 million.
SQ Magazine’s Takeaway
I think OpenAI is making a smart move here, not because Windows is new, but because Windows is where a huge chunk of real work gets done. The big win is not another chatbot that writes snippets. The win is coordination, running multiple agents safely, reviewing diffs when you want, and letting the boring repeat tasks happen automatically. If OpenAI can keep the sandboxing tight and the workflows predictable, Codex could become the kind of tool developers leave open all day, the same way we treat terminals and IDEs.
