What changed
OpenAI's September 29 release notes add Codex Cloud for Enterprise. Authorized members can share reusable cloud environments so teammates can start and continue Codex tasks from desktop, web, or mobile. Each task still receives its own isolated workspace, and cloud work can continue while the user's computer is asleep.
An environment is the prepared execution context around a project: repositories, tools, dependencies, access settings, credentials, and network configuration. The important product change is that this setup can now be published once and reused by an authorized team instead of being recreated by every person.
The rollout is described in OpenAI's ChatGPT Enterprise and Edu release notes.
Why it matters
This shifts Codex Cloud from an individual remote runner toward a shared execution layer for engineering teams. The collaboration unit is no longer just a prompt, repository, or chat transcript; it can also be a known-good environment that carries the project's tooling and access assumptions.
That separation is useful: the environment can be shared while each task keeps isolated working files and changes. A teammate can pick up the same prepared project context without inheriting someone else's unfinished task. It also makes long-running work more practical because the agent can keep working across devices and while a developer is offline.
There is a security boundary to keep clear. OpenAI's Codex Cloud guide says that sharing an environment exposes its prepared setup to authorized workspace members, but does not grant access to another person's task or automatically grant repository permissions. Environment-owned files, credentials, or connections can still expose shared resources, while personal vault values are not copied.
What builders should take away
For teams adopting hosted coding agents, treat the environment as a versioned operational artifact:
- Publish a small, reproducible baseline with the right dependencies, network policy, and least-privilege connections.
- Keep task state and source-control review separate: every task should remain independently inspectable, testable, and mergeable.
- Audit environment-owned credentials and connected accounts before sharing; environment sharing is not the same as repository authorization.
- Use the new model for repeatable team workflows, while reserving personal environments for experiments or sensitive setup.
The Codex plan and access guide explains that Codex Cloud uses reusable environments, runs tasks on OpenAI-managed computers, and keeps cloud access subject to workspace settings. The developer guide for cloud environments covers repository setup, dependencies, credentials, and network access in more detail.