Codex keeps coding in the cloud after your laptop closes

OpenAI made Codex’s cloud development environments reusable at DevDay on September 29, TechCrunch reported, so a project’s setup is prepared once and every later task starts from it, keeps working while your computer sleeps, and can be picked up from the web, mobile or desktop app. If your team runs many Codex tasks against the same repositories, setup stops being something each task redoes; faster starts and a shared workspace with approved settings and permissions are what OpenAI says the change is for.

Until now, each Codex cloud task was treated as an isolated remote sandbox. The environments now persist and can be configured, and OpenAI pitches Codex as something you run wherever you need it: on a computer, remotely from a phone, or in the cloud.

Publishing an environment

OpenAI’s Codex Cloud page lays out the setup. You choose repositories, and Codex inspects the project, prepares its dependencies and tools, and asks for any details or access it is missing. You review the checks before publishing the environment. Tasks then start from that published environment, each in its own workspace, so they share the setup but keep their working state separate. When the project changes, you update the setup instead of starting over.

Access to outside services sits in the environment settings. That is where you configure network access and supply credentials for the package registries, APIs and private services a project needs. For keys, OpenAI describes network secrets that work with allowed HTTPS services, next to a setting that limits which network destinations the environment can reach. A team that wants tasks to touch a private registry has to set that up per environment, before the first task runs.

Results come back as changed files and checks. From there you can request follow-up changes, then commit or open a pull request. The terminal works as an entry point too: the CLI starts a cloud task and lists recent cloud chats with their status.

The headline promise is one Anthropic already made when it took Claude Code’s cloud sessions out of research preview on its Pro, Max and Team plans: close the laptop and the agent keeps working. What OpenAI adds on top is the shared, published setup that every task inherits.

What else changed in Codex

• Codex CLI: tasks can be started and directed by voice, and a new /agents view tracks several delegated tasks at once. Editing prompts, resuming sessions and working with worktrees got smoother, and a cleaner terminal interface is meant to keep long sessions readable.
• Code review in the ChatGPT desktop app: summaries, a view of the code changes and questions to Codex about potential issues, before you leave feedback on GitHub pull requests or GitLab merge requests. Automatic reviews let Codex do a first pass while you are away from the computer.
• APIs: a new Decisions API uses Luna to answer a set of user-defined questions with predefined answers, and an updated Agents API adds computer use plus the option to build OpenAI agents with Amazon’s Bedrock Managed Agents that run entirely on AWS.

Security scans that run unattended

Codex Security Cloud is the piece that leans hardest on the persistent cloud. It scans entire GitHub repositories on demand, on a regular schedule, or whenever new commits arrive. Codex then investigates the findings, removes duplicates and prepares fixes in the cloud, even when the user isn’t active and the laptop is closed.

The toolset also includes access to the models from Daybreak Blue, OpenAI’s cybersecurity initiative, without a separate application.