Crashes, No Sound, or Screen Glitches?
Random freezes, missing sound and display glitches usually trace back to one bad driver. Find and replace yours safely.Free scan · under a minuteWindows Errors? Fix Them Before They Spread
Repair common Windows errors and clear accumulated junk for a smoother, more stable PC - no reinstall needed.Free scan · no reinstallOpenAI Codex is a coding workspace in the ChatGPT desktop app. It can work on a project to write or debug code, run tests and commands, review changes, and handle repository tasks. In local workflows, it can use folders, terminals, and developer tools, but its actual reach depends on the folder access, sandbox settings, approval rules, and workspace policies in effect. Local work does not mean that all messages and task context stay on your computer.
What Codex can do with your code and files
You can open a local folder or repository and give Codex a coding task. It can make changes to project files, investigate or debug code, run tests, execute other commands, and review proposed or completed changes. The app is an agentic workspace: rather than only suggesting text for you to copy, it can take actions in the project environment it has been given.
OpenAI describes editing the active folder or branch as the default scope. That is a default, not a universal guarantee about every installation: granted folder access, configured writable locations, app settings, and organization policy affect what Codex can access or change. Review the working folder and the proposed changes before accepting them, particularly in repositories with important or uncommitted work.
Can Codex run commands on your computer?
Yes. In local workflows, Codex can use a terminal to run commands as part of a task, including tests and other development commands. Whether a particular command can proceed without interruption depends on the sandbox and approval policy configured for that environment.
Recommended Free Tools
#1 Best Overall
The sandbox sets technical boundaries, such as which locations Codex can write to and whether it can use the network. Approval rules determine when it must ask before an action crosses a boundary or matches a configured rule. A prompt may request approval for one action or for a session. Command rules can permit routine commands while blocking or requiring approval for specified riskier patterns. In managed workspaces, administrators can enforce requirements that an individual user cannot override.
OpenAI’s app announcement describes the default posture as limiting agents to editing files in their working folder or branch and using cached web search, while asking permission for elevated actions such as commands requiring network access. The exact controls are configurable and installation-specific. A sandbox is a set of restrictions, not a promise that every action is harmless; inspect approval prompts and command effects rather than treating the word “sandboxed” as a blanket safety guarantee.
Codex Local and Codex Cloud are different execution paths
| Mode | Where coding work runs | What that means |
|---|---|---|
| Codex Local | Your desktop environment, through desktop, CLI, or IDE workflows | Codex works in the local project environment available to it, subject to folder access, sandbox limits, and applicable approval or workspace rules. |
| Codex Cloud | OpenAI-managed computers | Cloud tasks run on those computers and can continue while your own computer is asleep. |
“Local” identifies where the coding work runs; it does not establish that the related conversation or task context remains exclusively on the device. OpenAI says messages and task context may be stored in the cloud even when the work itself runs locally.
What happens to code and task data
Data handling depends on the account and agreement used with ChatGPT. OpenAI says the applicable terms and privacy policy, or the relevant enterprise, business, or API agreement, govern data shared through Codex. Its plan guidance says business-product inputs and outputs are not used to improve models by default. Pro and Plus conversations may be used unless training is turned off in data controls.
Those distinctions are why it is not accurate to promise that code is never transmitted or never used for model improvement. Check the data controls for your account and any organization-specific policy before sharing code you are not permitted to disclose. A local project folder can still be the work target while messages and task context involve cloud services.
Independent reader supportYour contribution helps us test, update, and keep practical guides available for everyone.What determines Codex’s actual access?
- Folder and repository access: The active folder or branch is the stated default editing scope, but additional writable roots or permissions may change the effective scope.
- Sandbox configuration: This defines technical boundaries for file writes and network access.
- Approval and command rules: These determine which actions can proceed, which trigger an approval request, and which are blocked.
- Workspace administration: Managed requirements can enforce restrictions beyond a user’s local preferences.
- Product setup: App version, platform, plan, rollout, and workspace controls can affect available features and behavior.
For an account-specific answer, inspect the current app and workspace controls rather than assuming that another user’s defaults apply. OpenAI distinguishes local and cloud workflows, and managed workspaces can control their access separately.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
