OpenAI’s Codex app has finally landed on Windows, and it’s every bit the full-featured release that developers on Microsoft‘s platform have been waiting for. The announcement came on March 4, 2026, via the official OpenAI Developers account on X, confirming that Windows users can now access the complete Codex experience — something that was previously exclusive to macOS since the desktop app’s debut in early February.

The backstory here matters. Codex didn’t start life as a sleek desktop app. It began as an open-source command-line tool — the Codex CLI — that developers had to run from their terminals. Then, in February 2026, OpenAI made a significant leap by shipping a standalone macOS app, complete with a multi-agent interface and the ability to coordinate multiple AI coding agents working in parallel on the same project. Windows developers were told it was “coming soon,” and roughly a month later, here it is.

So what exactly is the Codex app? At its core, it’s a collaborative workspace where AI agents — powered by tailored versions of ChatGPT — write, plan, and execute code based on your natural language instructions. You can give it something as simple as “list the contents of this directory” or as ambitious as “create an app that transcribes audio recordings,” and the agents will get to work. Before they do anything, they can strategize, laying out a detailed plan that you can review and comment on before giving the green light.​

The Windows release isn’t just a port, though. OpenAI actually built something new for it — a first-of-its-kind Windows-native agent sandbox. This is a big deal because, on other platforms, running AI agents securely typically requires Windows Subsystem for Linux (WSL) or a virtual machine. Now, Codex runs natively in PowerShell with OS-level controls including restricted tokens, filesystem ACLs, and dedicated sandbox users. The sandbox blocks agents from writing to files outside the working folder and prevents network access without your explicit approval. It’s a thoughtful security architecture, and it means developers can stay entirely within Windows-native workflows without having to spin up a Linux environment unless they specifically want to.

That said, flexibility is still there for those who prefer it. The Codex app on Windows also supports WSL2 — you can switch between the native Windows agent and a WSL-based one through the Settings menu, with the choice taking effect after a restart. You can even mix and match: run the agent in WSL while keeping PowerShell as your integrated terminal, or go fully native. The setup caters to the reality that many Windows developers live somewhere between both worlds.​

One of the standout aspects of Codex — regardless of platform — is the multi-agent capability. The left sidebar of the application lets you monitor multiple Codex conversations simultaneously, so different teams of AI agents can be working on separate projects at the same time. You’ll get notifications whenever an agent requires your approval for an action that needs elevated permissions, keeping you in the loop without constant micromanagement. There’s also support for automations to handle repetitive tasks like bug testing, and a “Skills” section that bundles together instructions, resources, and scripts, letting agents connect to specific tools and workflows.

Codex also integrates with both local directories and remote GitHub repositories. Agents can explore project branches before making any modifications, and you can set up worktrees — essentially sandbox environments where AI-generated work is kept isolated until you decide it’s ready for production. For teams, this is a genuinely useful guardrail that prevents experimental AI work from accidentally bleeding into live codebases.

Session continuity is another practical touch. Your conversation history is tied to your OpenAI account, so if you start a coding session on a Mac and then switch to a Windows machine, you pick up exactly where you left off. For developers who work across devices or operating systems, that’s the kind of seamless handoff that removes friction from the workflow.​

The app is powered in part by GPT-5.3-Codex, a model OpenAI introduced in early February 2026 specifically tailored for agentic coding tasks. OpenAI also described a lighter-weight companion model as a “daily productivity driver” suited for rapid prototyping, suggesting there’s a tiered approach to how compute is allocated depending on the complexity of the task at hand.

In terms of availability, Codex is accessible to ChatGPT Free, Go, Plus, and Pro users. That’s a broad base, though it’s worth noting that heavy, extended usage can eat through token allowances quickly — even on paid plans. Users exploring multi-agent workflows with complex projects will want to be mindful of that, especially if they’re running several parallel threads at once.

The competitive context is worth keeping in mind, too. Codex is going up against tools like Anthropic’s Claude-powered coding environments and a growing number of AI-native IDEs and coding agents. What distinguishes Codex, at least on paper, is the depth of the multi-agent orchestration and the security-first sandbox architecture — especially now that Windows has a native implementation. The fact that the Windows sandbox was purpose-built, using OS-level primitives rather than retrofitted from a macOS design, shows a level of platform investment that developers tend to notice and appreciate.​

For the Windows developer community — which remains massive and deeply embedded in enterprise, game development, and systems programming ecosystems — this arrival is more than just a checkbox. It brings one of the most talked-about AI coding platforms into their native environment, without compromise. Whether Codex becomes an essential part of the daily coding toolkit or remains a powerful but occasionally expensive novelty will depend on how developers integrate it into real workflows over the coming months. But today, at least, they finally have the chance to find out.