OpenAI’s Codex gets reusable cloud environments that follow developers across devices
What matters here is not just that Codex can keep a coding session alive in the cloud, but that OpenAI is moving AI-assisted development closer to a managed execution platform. For engineering teams, reusable environments can reduce one of the most error-prone parts of agentic coding: recreating dependencies, services and startup checks consistently enough for repeatable results. That has obvious appeal for onboarding, branch-specific work and long-running background tasks, but it also shifts attention from prompt quality to environment governance.
The practical question is whether these saved environments become an extension of the organization’s build and developer-platform strategy or a parallel path around it. Teams will need to decide how environment definitions are reviewed, who owns install scripts and health checks, and how repository drift is detected before an agent works from stale assumptions. If these environments are widely shared, they start to resemble lightweight golden images for AI-driven development, with the same lifecycle concerns around versioning, approval and rollback.
The security model also deserves scrutiny. Network secrets, allow-listed domains and per-user credential injection are useful controls, but they introduce new boundaries to audit: what outbound access is truly required, how secret use is logged, and whether shared environments can expose unintended supply-chain paths through dependency installation. Because each task runs in its own VM and persists independently, operations teams should also think about retention, task sprawl, cost visibility and incident response. In practice, the value of this feature will depend less on convenience than on whether it can be integrated into existing CI/CD, access-control and software supply-chain controls without creating a second, harder-to-govern developer environment stack.
OpenAI Group PBC today rolled out reusable cloud environments for its Codex coding agent, giving developers a saved project configuration they can work from across devices.
Announced at the company’s DevDay conference, the release targets the preparation that comes before an agent writes a line of code. A cloud environment does that work once. Each new task starts from that saved setup, with the project’s dependencies already installed.
Developers begin by connecting a repository from GitHub Inc. Codex reads the code to see which software versions the project relies on and drafts the install script from what it finds. Startup checks for the project’s services come out of the same pass, and OpenAI’s documentation says any changes are made in conversation with Codex.
Once an environment is published, each task launched from it gets its own virtual machine and keeps working while the developer’s computer sleeps. Changes not yet sent back to the repository stay with the task, so it can be reopened later from a phone or a browser without losing them. Dependency caches survive the background refreshes that keep the repository current.
OpenAI is also pitching the environments at teams that want every developer working from the same approved configuration. After one is shared with a ChatGPT workspace, colleagues can run their own tasks from it. Access to a shared environment does not extend to anyone else’s tasks or to editing the setup. Enterprise administrators hand out that editing permission separately.
For the credentials a project needs, cloud environments support standard environment variables and a second type OpenAI calls network secrets. A program that uses a network secret sees only a placeholder, with a proxy inserting the real value as traffic heads to a domain on the allow list. Individual users of a shared environment can be asked to supply their own values from a personal account. Internet access defaults to a preset of common package registries, and developers can open it wider.
Codex has handled tasks in the cloud since it debuted as a research preview inside ChatGPT in May 2025. Codex Security Cloud, another DevDay launch, takes the same approach to security work. It scans GitHub repositories on a schedule and prepares fixes while the developer’s machine is off.
Access to cloud environments is rolling out to ChatGPT Plus and Pro subscribers, and Business, Enterprise, Edu and Healthcare workspaces get it too. Plus accounts run on the smallest virtual machines, with half the processors and memory of the Pro, Business and Enterprise tiers.
By default, a task’s state can be recovered for up to seven days after it was last used. Codex cannot yet work with repositories on GitLab Inc. or on self-hosted GitHub Enterprise Server, and computer and browser use are also unsupported in cloud environments.
OpenAI’s Codex gets reusable cloud environments that follow developers across devices
Enjoyed this article? Sign up for our newsletter to receive regular insights and stay connected.