On September 29, 2026, OpenAI announced reusable Codex Cloud environments: developers can prepare a setup once, then use it to start new coding tasks. Each task gets its own isolated workspace, and eligible users can continue working across desktop, web, and mobile.
OpenAI announced reusable Codex Cloud environments
A Codex Cloud environment is a prepared setup of repositories, dependencies, tools, and access settings for coding tasks. Reusing that setup means teams can start new tasks without rebuilding the same foundation each time.
Reusable setup, isolated tasks
The environment is the starting configuration, not a shared working directory. Each new task receives a separate workspace based on the published setup, while existing tasks keep their own files and changes. Developers can start or continue tasks from desktop, web, or mobile, and a cloud task can keep running while their computer is asleep.
OpenAI’s demonstration walks through selecting repositories, configuring an environment, and returning to a task from a phone after closing a laptop.
Saved task state is recoverable for up to seven days
Codex Cloud task state is recoverable for up to seven days after the last turn or resume. That recovery period applies to saved task state; each task still has its own isolated workspace.
Plans, billing, and task resources
OpenAI lists Codex Cloud for eligible Plus, all Pro tiers, Business, Enterprise, Healthcare, and Education accounts. Free and Go are excluded. Access is subject to rollout and workspace settings.
Standard cloud environments have no separate VM charge at launch. Model usage still counts toward the account’s normal Codex limits and applicable credits or billing.
OpenAI lists these default virtual-machine resources per cloud task:
| Plan group | vCPU per task | Memory | Disk |
| Plus, Edu Plus | 2 | 8 GiB | 8 GiB |
| Pro, Business, Enterprise | 4 | 16 GiB | 32 GiB |
| Edu, Edu Pro | 4 | 16 GiB | 32 GiB |
Enterprise customers can request larger virtual machines or custom configurations through their OpenAI account team.
Configure the setup and credentials
Environment creation and publishing are available on web or desktop. Developers select GitHub repositories, let Codex prepare and test the setup, review the configuration, and publish it. They can configure installation commands and a Start skill; changes to a setup must be saved and republished before new tasks use them.
Environment variables are passed directly to programs running in the workspace. For network secrets, programs receive placeholders, and OpenAI’s proxy substitutes credentials for allowed HTTPS destinations on port 443. A shared environment can also request credentials that individual users provide from personal vaults.
Cloud environment limits and GitLab
OpenAI lists computer and browser use, GitLab repositories, and self-hosted GitHub Enterprise Server as unsupported in Codex Cloud environments. GitLab merge-request review belongs to the separate Code Review capability, not the Cloud environment workflow.
