Codex in the Cloud: OpenAI’s Coding Agent Now Works While Your Laptop Is Closed

Here is a small change that alters how developers work: your coding agent no longer stops when you close your laptop. At DevDay on 29 September 2026, OpenAI gave its software-engineering agent, Codex, reusable cloud environments — workspaces that live in the cloud, keep your setup, and keep running whether you are at your desk, on your phone, or nowhere near a computer at all. If you have ever left a long task running overnight and hoped your laptop would not sleep, you already understand why this matters.

What a reusable cloud environment actually is

Until now, Codex cloud tasks mostly worked like this: you asked for something, Codex spun up a fresh remote sandbox, did the job, and the sandbox went away. The new environments are different because they persist. You set one up once with your repository, tools, dependencies, permissions, and settings, and future tasks start from that same setup instead of starting cold every time.

That has two practical effects. Tasks start faster, because the environment is already warm. And teams can share one approved setup, so everyone works from the same tools and permissions instead of each person configuring their own machine. The environments are isolated per task, so one job cannot quietly interfere with another.

And yes — you can open them from anywhere. OpenAI says the same environment is reachable from desktop, web, and mobile. Start a task at your desk, check it from your phone on the way home, and pick it up again tomorrow without losing your place.

The refreshed command line

The Codex CLI got a solid refresh in the same update:

• Voice commands: you can now start and steer tasks by speaking, which suits hands-busy moments and accessibility needs alike.
• A new /agents view: delegate work to several agents at once and track them in one place — one agent can hunt a bug while another works on a feature and a third reviews changes.
• Everyday workflow fixes: better prompt editing, session resumption, worktree support, and a cleaner terminal interface for long sessions.

The refreshed CLI is available across all plans, which makes it the easiest part of this update to try.

Code review inside ChatGPT

Codex also moved into a new code review experience inside the ChatGPT desktop app. You can read summaries of changes, explore diffs, and ask Codex questions about a pull request before you leave feedback on GitHub or GitLab. Automatic reviews can run in the cloud while you are away, so the first pass is done by the time you sit down.

GitHub review is generally available; GitLab support is in preview. For a solo developer, the value is a second pair of eyes that never gets tired. For a team, it is a faster first review, though a human should still make the final call on anything that touches production.

Security Cloud: scans that run on a schedule

The other big piece is Codex Security Cloud. It scans whole GitHub repositories on demand or on a schedule, checks new commits as they land, removes duplicate findings, and prepares fixes for review. Because it runs in the cloud, it keeps working with your laptop closed — a scheduled scan does not care where you are.

It also includes access to OpenAI’s Daybreak Blue cybersecurity models without a separate application. That matters because security scanning used to mean bolting on yet another tool with yet another signup. Here it sits inside the same system that already knows your code.

Who gets what

Availability breaks down by plan:

• Codex in the cloud (reusable environments, cross-device access): Plus, Pro, Business, Healthcare, Edu, and Enterprise.
• Refreshed CLI and code review: all plans.
• Codex Security Cloud: Pro, Business, Enterprise, and Edu.

If you are on a free plan, you can still use the refreshed CLI, but the cloud environments and security scanning need a paid plan.

What this changes for a small team

Think about a two-person team, or a solo developer with a side project. Before this update, serious agent-assisted coding meant keeping a machine awake, keeping environments in sync, and doing security checks by hand or with extra tools. Now the pattern looks more like this: set up one cloud environment per project, hand Codex a task, let it run while you do something else, and review the result — including a security pass — when it is ready.

That does not replace good engineering habits. Codex still needs clear tasks, and its proposed fixes still deserve a human review before they ship. What it does replace is the waiting: the dead time between asking for work and seeing a result, and the setup tax of rebuilding an environment for every job. For developers who already use Codex, this is the update that makes it feel less like a tool you babysit and more like a colleague you hand work to.