Markdown edition

# Set up a Codex Cloud environment for a repository

> Connect a GitHub repository, publish a reusable Codex Cloud environment and learn how isolated tasks continue across desktop, web and mobile. This documented guide explains access, network, credentials and cost; BIG CHANGE has not tested it.

By BIG CHANGE Editorial

Published: 2026-09-29T21:39:48.290Z
Updated: 2026-09-29T21:39:48.290Z
Canonical: https://bigchange.ai/blog/codex-cloud-environment-setup-guide

![A person seated at a desk using an unbranded laptop beside a window at sunset](https://bigchange.ai/api/media/file/codex-cloud-workspace-hero-v2.png)
AI-generated by OpenAI; BIG CHANGE editorial illustration.

**Summary:** Connect a GitHub repository, prepare and publish its Codex Cloud environment, then start a coding task you can reopen on desktop, web or mobile. This guide follows OpenAI’s documentation as of September 29, 2026; BIG CHANGE has not tested the workflow.

Codex Cloud runs coding tasks on OpenAI-managed computers. Its September 29 release notes say tasks can continue while your own computer is asleep. The setup has two parts: an **environment** holds the repository, tools, dependencies and access settings you want to reuse; a **task** gets its own isolated workspace from the published environment. Returning to the same task preserves its working files, while starting another creates separate work. [OpenAI release notes](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes); [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).

## The big change

- **What changed:** OpenAI now documents a reusable cloud setup for a repository that can launch separate coding tasks on its managed computers. A task can continue while your computer is asleep and can be reopened across desktop, web and mobile. [OpenAI release notes](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes).
- **Why it matters:** You can prepare dependencies and tools once, then start new tasks from the published environment. Reopening an existing task continues its own saved work; creating another starts an isolated workspace. The guide below shows the setup and return path. [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).
- **What to watch:** Publishing updates the starting point for new tasks, not tasks already underway. Review repository access, credentials and allowed network destinations before publication. OpenAI documents the controls and their limits; BIG CHANGE has not independently run this workflow. [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).

## Before you begin

Sign in with a ChatGPT account eligible for Codex Cloud and have a GitHub repository you can connect. OpenAI lists eligible Plus, Pro, Business, Enterprise, Healthcare and Education accounts, subject to rollout and workspace settings. Free and Go include Codex but not Codex Cloud. Guest, K–12 and Enterprise view-only seats cannot create cloud environments. Enterprise cloud access can be disabled by an administrator; creating a shared workspace environment needs an additional permission. [Using Codex with your ChatGPT plan](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan).

Create and publish the environment on the web or in the ChatGPT desktop app. Mobile can use an available environment after it has been published. No API key or separate API setup is part of this documented ChatGPT account workflow. [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).

## Connect the repository and publish the setup

1. In a new Codex task on web or desktop, choose **Work in > Cloud**, open **Select environment**, then choose **Create environment**. You can also use **Settings > Codex Cloud > Environments > Create environment**.
2. Select the GitHub repository or repositories to check out. Connect GitHub if prompted, then select **Get started**.
3. Codex inspects the repository and prepares dependencies and tools. Supply missing information or access when asked. You can specify versions, commands and services your project needs.
4. Review its setup report, configuration and files. Resolve unfinished setup work, save changes, then select **Publish**. Wait for **Environment published** before starting a task. Saving stores configuration; publishing captures the prepared filesystem for new tasks. [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).

Check the environment’s network and credential settings before publishing. An environment variable is passed directly to programs that need its value. A network secret is intended for a credential sent to specified HTTPS destinations: programs receive a placeholder and a proxy substitutes the credential for allowed requests. Internet access also has a separate domain policy. Allowing a domain does not supply credentials or grant permission to that service. OpenAI says Enterprise security requirements apply alongside the environment’s settings. These choices determine what the cloud task can reach, so give the environment only the access its work requires. [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).

## Start a task and return to it

After publication, select **Start a new task**, describe the repository work and send the request. OpenAI documents that Codex can edit files, run commands and check its work in the cloud workspace; review its changes and results before committing or opening a pull request. [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).

On the web, use **Work in > Cloud** and select the published environment. On mobile, open **Codex** and choose an available environment. To continue work already underway, **reopen the same task**. Starting a new task from that environment starts separate work. OpenAI says an existing task keeps its own saved files, including uncommitted changes and installed tools, and that saved VM state is recoverable by default for up to seven days after the last turn or resume. Commit important work or otherwise save output you need; task state does not replace source control. [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).

If the reusable setup needs a change, edit it under **Settings > Codex Cloud > Environments**, let Codex prepare and test the revision, save it and select **Republish**. New tasks use the updated setup; existing tasks keep their own state. [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).

## What this cloud task can reach

Codex Cloud works in its cloud workspace. It is distinct from remote access to a task running on your computer, which still depends on that computer and its access settings. OpenAI’s current environment guide lists computer and browser use as unsupported in cloud environments. The steps above do not give the task access to files on your local machine; such access would require a separately documented, connected feature. [Using Codex with your ChatGPT plan](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan); [Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments).

At launch, OpenAI says standard Codex Cloud environments have no separate VM charge. Model use counts toward normal Codex limits and any applicable credits or billing. Check your account’s usage display and [current pricing](https://learn.chatgpt.com/docs/environments/cloud-environments) before a long task; the cost depends on your plan and usage. [Using Codex with your ChatGPT plan](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan).

## Sources & further reading

- [ChatGPT Business release notes, September 29, 2026](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes) — Announces Codex Cloud’s cross-device use, reusable environments, isolated task workspaces and ability to continue while a computer sleeps. It does not provide the full setup procedure.
- [OpenAI’s Cloud environments guide](https://learn.chatgpt.com/docs/environments/cloud-environments) — Documents the repository connection and publication steps, task reuse, secrets, network controls, return path, saved state and current limitations. Interface labels and limits may change.
- [Using Codex with your ChatGPT plan](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan) — Establishes plan eligibility, workspace controls and the launch pricing rule: no separate standard VM charge, with model use charged against applicable Codex usage or billing.

## Sources

- [ChatGPT Business release notes, September 29, 2026](https://help.openai.com/en/articles/11391654-chatgpt-business-release-notes) — Dated September 29 release note for Codex Cloud; announces reusable repository/tool/dependency environments, isolated per-task workspaces, desktop/web/mobile use and tasks that continue while the local computer sleeps. It does not provide setup details.
- [Codex Cloud environments](https://learn.chatgpt.com/docs/environments/cloud-environments) — Official current guide; documents repository connection, setup, publish/republish, task continuation, isolation, network settings, credentials, access, seven-day saved VM recovery and unsupported cloud features. Vendor documentation, not independent validation.
- [Using Codex with your ChatGPT plan](https://help.openai.com/en/articles/11369540-using-codex-with-your-chatgpt-plan) — Current availability and use/cost details. Eligible Plus, Pro, Business, Enterprise, Healthcare and Education accounts subject to rollout and workspace controls; not Free/Go; no separate standard VM charge at launch, model usage draws on applicable plan allowances/credits/billing.
- [OpenAI DevDay 2026](https://devday.openai.com/) — Confirms event on September 29, 2026, location, timetable and public keynote availability; does not list product specifications.
- [OpenAI livestreams](https://openai.com/live/) — At 2026-09-29 check, the livestream archive did not list a DevDay 2026 replay. No claim about an unavailable replay beyond this check.