GuidesGetting started

“Please contact your workspace admin to enable device code authentication” — the fix

A Codex sign-in error asking for your workspace admin is a permission issue. What the admin enables, how to retry, and supported alternatives if it stays off.

August 10, 2026Updated September 13, 2026The Everpod team
The short answer

A device-code error telling you to contact your ChatGPT workspace admin points to a workspace permission, not a mistake in the code you entered. Ask an admin to enable device-code login in the workspace permissions, then start a fresh sign-in. Changing your personal Security setting does not grant a permission the workspace controls. OpenAI documents the two settings in its headless authentication guide.

Why the same account can work on a laptop

The ordinary browser login and the device-code flow are different sign-in methods. Codex CLI normally opens a browser and waits for a localhost callback. Device-code login lets you approve the terminal’s sign-in from another browser, which is useful on a server or when that callback is blocked.

Success with the first method does not enable the second. Buying a different ChatGPT plan or repeatedly requesting new codes does not resolve a workspace permission refusal. If you intended to use a personal account, first check which account and workspace the browser selected.

What your admin needs to know

Ask for device-code authentication for Codex in the workspace permissions, explain which tool and machine you need to sign into, and include the error text. The request is to permit a sign-in method, not to increase your model allowance. An admin should also check that your workspace role permits Codex access.

Once the permission is enabled, run a new attempt on the machine that needs access. For Codex CLI:

codex login --device-auth
Copied

For OpenClaw, use its own OpenAI device-code sign-in:

openclaw models auth login --provider openai --device-code
Copied

Open the printed link, use the intended ChatGPT account and approve that attempt’s code. Wait for the terminal to confirm completion. A code left open while permission was being arranged may have expired.

If the workspace keeps device-code login disabled

Ask which supported method your organization allows. OpenAI’s Codex CLI documentation includes a browser login with its callback forwarded over SSH, and an authenticated-cache transfer to a trusted headless machine. Those are alternatives to the device-code flow, not authorization to disregard a workspace policy. Credential transfer needs the same care as handling a password.

An approved API key is another option. It uses separate API billing rather than the ChatGPT subscription allowance, so confirm the organization’s budget and access controls before switching. Codex CLI’s fallback instructions should not be assumed to configure OpenClaw as well; its subscription connection has its own sign-in and model selection.

If permission is enabled but access still fails

Check the browser’s selected account and workspace, then distinguish a login failure from a model-access error. For Codex CLI, codex login status checks whether a login is active; a small task checks that it can actually run. For a personal account that is refusing device codes without a workspace involved, use the personal Security-settings guide.

Give your AI agents a permanent cloud computer.

We set it up and secure it, so all you do is use them. Run Claude Code, Codex or any agent you use, more at once than your laptop can hold, and keep them working when it’s closed. Or have OpenClaw set up and run for you.

See the developer pod

From $24/mo · your own plan or API key

Get your OpenClaw agent

$29/mo, cancel anytime · model usage included