Can you use your ChatGPT subscription in Claude Code?

No, your ChatGPT subscription is not a native Claude Code login. Use Codex with ChatGPT sign-in for OpenAI subscription access, or use an interface such as T3 Code to control Claude Code and Codex as separate agents.

Checked on 26 October 2026. The useful distinction is between the application window, the agent doing the work and the account paying for its requests. Sharing a workspace does not make those accounts interchangeable.

Why does ChatGPT sign-in not appear in Claude Code?

Anthropic’s authentication documentation lists Claude accounts, Claude Console and supported infrastructure routes. ChatGPT is not a native account option. OpenAI’s authentication page puts ChatGPT subscription sign-in in Codex instead.

That supports a narrow answer: there is no documented native route for applying a ChatGPT subscription to Claude Code. It does not mean you cannot use the two agents on the same repository. Nor does it mean every product with a Claude-style terminal interface is the official Claude Code.

What you want	Documented starting route	Account check
OpenAI coding agent on a ChatGPT plan	Codex CLI or IDE extension, signed in with ChatGPT	Confirm subscription sign-in rather than API-key authentication
Native Claude Code	Claude account or another route in Anthropic’s authentication docs	Confirm the intended Claude account or provider
One control surface for both agents	T3 Code, with each agent configured first	Select the runtime as well as the model
A workspace app around configured agents	Conductor	Check its authentication settings before starting work

For other subscription-compatible tools, compare the ChatGPT hub and Claude hub. They describe separate routes, not a shared credit balance.

How do you use Codex on the same project instead?

Install the official Codex CLI through OpenAI’s setup documentation, then run codex login and choose ChatGPT sign-in. OpenAI documents codex login status for checking the active authentication method. An API-key login is a different billing choice.

Our suggested first task is read-only: ask Codex to identify the relevant files and propose a test plan. Confirm the repository and account before allowing edits. This is a recommended safety check, not a claim that we benchmarked the two agents.

If Claude Code already made changes, save a handoff containing the intended outcome, modified files, test results and remaining questions. Review or commit those edits before another agent changes the same files. Do not assume that a login change imports conversation history, permissions or approvals.

The general setup guide covers the broader subscription-versus-key distinction. If the editor is Cursor, use the separate ChatGPT-in-Cursor article, rather than treating Cursor’s agent and its extensions as one product.

Can one app switch between Claude Code and Codex?

Yes. The T3 Code README describes control of agents already configured on your computer. It explicitly gives separate installation and login prerequisites for Codex and Claude Code. Selecting Codex is selecting another agent, not turning Claude Code into an OpenAI client.

Conductor’s FAQ says it uses authentication already saved on your machine by default and offers API-key overrides. Inspect the chosen harness and account route rather than assuming the app name determines the bill.

This is the practical option if what you like is a workspace, review panel or remote interface. This page addresses which subscription can authenticate which runtime, rather than parallel-agent coordination.

Does adding an OpenAI key make the work count toward ChatGPT?

No. OpenAI’s authentication documentation separates ChatGPT subscription access from API-key usage at API rates. An OpenAI-compatible endpoint is not evidence that ChatGPT subscription usage is supported.

Treat a tutorial promising a drop-in subscription proxy as a separate integration requiring verification. Check who maintains it, which credentials it handles and whether the provider documents the route. Do not copy login files into an unknown bridge. Anthropic’s legal page also places conditions on products running Claude Code, including keeping its binary unmodified and letting users authenticate for themselves.

What should you check before handing work between agents?

Use this short checklist: identify the agent, confirm its authentication method, inspect its working directory, review pending changes and give it one bounded task. Keep paid fallback routes explicit. If either agent reaches a limit, moving to another runtime is a new account choice, not a way to enlarge the first account’s allowance.

For the basic terminology, see what bring your own subscription means. The shortest decision is still: choose Codex for your ChatGPT sign-in, and choose Claude Code for a supported Claude Code authentication route.