Codex for Windows Gets Computer Use and Mobile Remote Control, Plus GitHub Enterprise Server Integration
ChatGPT Enterprise's Codex app now supports Computer Use on Windows with iOS/Android remote handoff, and a new GitHub Enterprise Server template for on-prem code review.
OpenAI has pushed a significant update to the Codex app for ChatGPT Enterprise and Edu workspaces. Two features that were previously Mac-only, Computer Use and mobile remote control, are now available on Windows. At the same time, a new GitHub Enterprise Server app template means teams running on-premises GitHub can connect their repos directly to Codex for code review and security scanning.
Here is what each of those things actually does, and why it matters.
Computer Use on Windows: Codex Can Now Operate Your Desktop
Computer Use lets Codex see your screen and interact with it. It can click buttons, fill in fields, navigate menus, and type, all inside native Windows applications. This is not browser automation or an API integration. It works with anything visible on screen.
That makes it useful in situations where no API exists. Internal tools built years ago, legacy admin panels, Electron apps, installers, multi-app workflows, even reproducing a bug that only surfaces through the graphical interface. If a human can click through it, Codex can work through it.
To use it, install the Computer Use plugin from the Plugins marketplace and invoke it with an @ mention in Codex. You can also let Codex decide when to reach for it based on the task you give it.
A few things to note before you get started. Computer Use on Windows requires early access, so you will need to contact your OpenAI account representative to be enrolled. It is also not available in the European Economic Area, the UK, or Switzerland at launch.
Remote Handoff: Control a Windows Machine from Your Phone
The second Windows addition is remote control. Once Codex is running a task on your Windows machine, you can pick up your iPhone, Android phone, or a Mac and continue directing it from there.
The architecture is straightforward. Your Windows PC stays the working host. It holds your project files, runs the shell, serves the local app, and maintains all the local context Codex needs. Your phone becomes the control surface. You send prompts, approve actions, respond to Codex when it needs guidance, and check progress without being at your desk.
This is particularly useful for longer-running tasks. You kick something off before a meeting, and you can steer it or approve the next step from your phone without the work stopping.
One important limitation: Windows can act as a host for remote sessions, but you cannot currently use Windows as the controller of another machine. Remote control is one-directional for now.
Admins can turn off locked computer use (where Codex continues working after the machine locks) by setting remote_computer_use = false in Policies and Configurations in Codex Cloud.
GitHub Enterprise Server: On-Prem Repos Now Connect to Codex
The third update in this release targets teams running GitHub Enterprise Server, the self-hosted version of GitHub. Until now, Codex integrations for code review and security scanning worked with GitHub Cloud. If your organisation keeps repos on-premises for data residency or compliance reasons, that integration was not available to you.
The new GitHub Enterprise Server app template changes that. Workspace admins can follow a guided setup flow to configure a GitHub Enterprise app, covering OAuth credentials, callback URLs, webhook details, and managed MCP server URLs. Once published, members get access to the workspace-specific connector for Codex Web, Code Review, and Security Review.
After publishing, the app is managed from Workspace settings, where admins can control role access, action permissions, and whether certain actions require confirmation.
What Codex Code Review and Security Review Actually Do
Since this integration brings both capabilities to on-prem repos for the first time, it is worth being clear about what each one does.
Codex Code Review integrates with GitHub pull requests. When a PR opens, Codex reviews the diff, checks it against any guidelines you have set in an AGENTS.md file at the root of your repository, and posts a standard GitHub code review. It flags P0 and P1 issues only, so the signal stays high and review comments do not become noise. You can define your own rules in AGENTS.md, for example instructing Codex not to log PII, and it will follow them.
Codex Security works differently. It functions more like a security researcher than a traditional scanner. It reads your code, builds a threat model specific to your codebase, scans repository history, and then attempts to reproduce potential vulnerabilities in an isolated sandbox environment to confirm they are real before surfacing them. For validated issues, it produces a minimal patch suggestion targeting the root cause. That patch does not get applied automatically. It comes up for human review and can be turned into a pull request through your normal workflow.
For Enterprise and Edu workspaces, access to Codex Security is managed through workspace permissions and supports RBAC, including SCIM-synced groups.
What This Means for Your Team
If your engineering team runs on Windows, they can now use Codex the same way Mac users have been able to since the Computer Use features launched earlier this year. Multiple agents running in parallel, isolated worktrees, reviewable diffs, all from a Windows desktop. The CLI and IDE integrations remain interoperable.
If your organisation keeps GitHub repositories on-premises and has avoided cloud-based AI tooling because of where your code lives, the Enterprise Server template gives you a direct path to bringing Codex into your review workflow without moving your repos anywhere.
The combination of automated, high-signal code review and sandbox-validated security scanning is a meaningful addition to a PR process, particularly for teams where security review is a bottleneck or where reviewers are spread thin.
The remote control capability is more of a quality-of-life improvement for individual developers, but for anyone running longer agentic tasks it reduces the need to stay at your desk waiting for Codex to need input.
For more detail on the GitHub integration setup, the OpenAI developers documentation covers the configuration steps. For Codex Security, the help centre article walks through access requirements and how the scanning process works.