Use Your Claude Pro Subscription in a Third-Party App? 2026 Rules

You can use a Claude Pro subscription in a third-party app only when the app runs Claude Code itself, unmodified, under your own login. As of September 2026, Anthropic, OpenAI and xAI let you sign in to their own coding CLIs (Claude Code, Codex, Grok Build) with a paid plan, and an app that runs one of those CLIs unmodified under your login is using your plan the way the provider built it. Anthropic and Google do not allow other software to take your plan's login and call the model directly, and Google's consumer plans no longer work in Gemini CLI at all. OpenAI is the most permissive. An API key works everywhere, billed per token.

This page is about what is allowed, vendor by vendor, and how to tell whether an app is on the right side of it. Whether a plan or a key is cheaper for your work is a separate question, covered in subscription vs API key: cost, limits and when each is right.

Checked against each provider's own pages on 15 September 2026. These rules changed several times this year, so re-read the sources before you rely on them. Nothing here is legal advice, and nothing here is an endorsement from Anthropic, OpenAI, Google or xAI.

Can you use a Claude Pro subscription in a third-party app? Vendor by vendor#

Two notes on that table. The OpenAI line about external tools is quoted from OpenClaw's documentation, not from an OpenAI page. OpenAI's own Codex README does recommend signing in with ChatGPT "to use Codex as part of your Plus, Pro, Business, Edu, or Enterprise plan". And "not addressed" means we found no page saying either way. It does not mean yes.

What changed in 2026?#

Most confusion comes from posts describing rules that have since moved, or a change that never took effect. Newest first:

• June 18. Gemini CLI stops serving Google AI Pro, Ultra and free individual accounts. Individuals move to Antigravity CLI; Code Assist Standard and Enterprise licences and API keys are unaffected.
• June 15. Anthropic pauses the Agent SDK change on the day it was due. Its help article now opens: "For now, nothing has changed: Claude Agent SDK, claude -p, and third-party app usage still draw from your subscription's usage limits." The promised credit "isn't available", and Anthropic says it will share an update "before anything takes effect".
• May 25. xAI launches Grok Build for SuperGrok and X Premium Plus subscribers.
• May 19. Google announces the Gemini CLI move to Antigravity CLI.
• May 13. Anthropic announces that from June 15, Agent SDK usage, claude -p and third-party apps built on the SDK would move to a separate monthly credit: $20 for Pro, $100 for Max 5x, $200 for Max 20x.
• Early May. Sam Altman posts "you can sign in to openclaw with your chatgpt account now and use your subscription there!"
• April 4. Anthropic tells subscribers that third-party harnesses such as OpenClaw will no longer use Claude plan limits. Boris Cherny, head of Claude Code, said people could keep using them "with your Claude login via extra usage bundles … or with a Claude API key."
• February 19. Anthropic's Claude Code legal page gains an "Authentication and credential use" section. As quoted in GitHub issues that week, it said plan OAuth "is intended exclusively for Claude Code and Claude.ai" and that using those tokens "in any other product, tool, or service — including the Agent SDK — is not permitted". The "ban" headlines come from that wording, which has since been rewritten.

What does Anthropic allow with a Claude Pro or Max plan?#

Anthropic's rules live on the Claude Code "Legal and compliance" page, under "Authentication and credential use". They are the strictest of the four and the most misreported.

What is restricted is developers, not you. The page says subscription OAuth "is intended exclusively for purchasers of Claude Free, Pro, Max, Team, and Enterprise subscription plans and is designed to support ordinary use of Claude Code and other native Anthropic applications." Then:

"Anthropic does not permit third-party developers to offer Claude.ai login into their own applications, or to route requests through Free, Pro, or Max plan credentials on behalf of their users. Moreover, developers may not collect, store, or intermediate Claude.ai credentials or session tokens — sign-in to a Claude account must complete through Anthropic's own flow."

What is still fine. The same page says the rules do not "prevent an end user from signing in to the unmodified Claude Code binary with their own Claude subscription, including where a platform hosts Claude Code." A product that runs Claude Code must not modify the binary or "remove, disable, or restrict any authentication method built into it", and "each end user must authenticate with their own Anthropic API key, Claude subscription plan credentials, or 3P inference provider credential."

The Agent SDK is the line people cross. Anthropic's Agent SDK overview says: "Unless previously approved, Anthropic does not allow third party developers to offer claude.ai login or rate limits for their products, including agents built on the Claude Agent SDK." An app with its own "Sign in with Claude" button built on the SDK is on the wrong side of that.

Two statements that are hard to square. The April 4 change bills third-party harnesses that use your Claude login as extra usage. The June notice says third-party app usage still draws from plan limits. If a tool warns that your Claude login is drawing extra usage, believe it, and check the usage page in your Claude settings.

Enforcement. "Anthropic reserves the right to take measures to enforce these restrictions and may do so without prior notice." The page also says "Advertised usage limits for Pro and Max plans assume ordinary, individual usage of Claude Code and the Agent SDK." Anthropic's Consumer Terms (effective October 8, 2025) still prohibit access "through automated or non-human means, whether through a bot, script, or otherwise", except via an API key or where Anthropic explicitly permits it; the legal page is where it spells out what it permits.

What does OpenAI allow with ChatGPT Plus or Pro?#

OpenAI's model is simpler. Codex is included in ChatGPT plans, and you sign in to it with your ChatGPT account.

• Plans. As of September 2026, OpenAI's Codex pricing docs say Codex is included in "ChatGPT Free, Go, Plus, Pro, Business, Edu, or Enterprise".
• Sign-in vs key. The Codex auth docs describe both. With ChatGPT sign-in, usage follows your ChatGPT workspace settings. With an API key, "OpenAI bills API key usage through your OpenAI Platform account at standard API rates."
• Automation. "Use API key authentication for programmatic Codex CLI workflows, such as CI/CD jobs."
• Third-party tools. OpenClaw's docs say "OpenAI explicitly supports subscription OAuth usage in external tools and workflows like OpenClaw," and Sam Altman said the same publicly in May. We found no written OpenAI policy page equivalent to Anthropic's. Codex being open source also means other apps can run it as published.
• Your token file. The Codex auth docs say the cached ~/.codex/auth.json contains access tokens and should be treated like a password: not committed, not pasted into tickets, not shared in chat.

So, does ChatGPT Plus include API access? No. The API is a separate account and a separate bill. What Plus gives you is Codex, which is often what people actually want the API for.

What does Google allow with a Gemini plan?#

Google announced on May 19 that "on June 18, 2026, Gemini CLI and Gemini Code Assist IDE extensions will stop serving requests for Google AI Pro and Ultra, as well as those using it free of charge." A Google AI Pro or Ultra plan now pays for Google's own Antigravity tools. Gemini CLI still works with a Gemini API key, a Vertex key, or a Gemini Code Assist Standard or Enterprise licence.

The third-party rule is blunt: "Directly accessing the services powering Gemini CLI (for example, the Gemini Code Assist service) using third-party software, tools, or services (for example, using OpenClaw with Gemini CLI OAuth) is a violation of applicable terms and policies." Google says this may lead to suspension or termination. So no third-party app can honestly offer you Gemini on a consumer Google plan.

What does xAI allow with a Grok plan?#

xAI's Grok Build CLI signs in through a browser login with grok login, and its authentication docs use XAI_API_KEY when no session token is active and recommend a key for CI/CD. We found no xAI page saying which consumer plans that login covers beyond SuperGrok and X Premium Plus, or what other apps may do with it. Use an xAI API key in any other app.

Three ways an app reaches your plan. Which are sanctioned?#

Every app that says "use your existing subscription" does one of these, or asks for a key instead. Sort an app into a row before you connect anything.

A quick test: with route 1, the provider's CLI is actually installed and its own commands still work (/status in Claude Code, /status in Codex). With routes 2 and 3, the model is reached some other way, and the CLI is either absent or only used as a source of tokens.

The lines are not always clean. An app can run the real CLI and still manage your sign-in for you, so ask two questions: is the provider's own CLI doing the work, and does your sign-in finish on the provider's own page? We don't recommend route 3 under any framing.

Does a third-party app get its own usage limits?#

No. An app that runs on your plan draws from the same pool as the vendor's own apps. Anthropic's help center says Pro and Max limits "are shared across Claude and Claude Code, meaning all activity in both tools counts against the same usage limits." OpenAI's Codex pricing docs say "Local messages and cloud chats share your plan's usage allowance." A heavy afternoon in any app leaves less for everything else.

How those limits work, how to watch them, and when paying per token is the better deal are in subscription vs API key: cost, limits and when each is right.

Where Universe fits#

Universe is a Mac app that takes route 1, so here is exactly what it does. Check it against the provider pages above and decide for yourself.

• It runs the providers' CLIs on your Mac. A copy of Claude Code and of Codex ships inside the app. Gemini CLI and Grok are installed on demand. Turns go from those CLIs on your machine to the provider, under your account.
• Claude sign-in uses Claude Code's own command. "Sign in with Claude" runs claude setup-token, which Anthropic's docs describe as opening "the same browser authorization flow as /login". The resulting credential is sealed in macOS secure storage on your Mac. Anthropic's page also says developers may not "store" Claude.ai session tokens, so weigh that sentence yourself. If you would rather not, give Universe an Anthropic API key instead.
• What pays. Claude Code runs on your Claude Pro or Max plan, or an Anthropic API key. Codex runs on your ChatGPT Plus, Pro or Business plan, or an API key. Gemini CLI runs on your Google account or a Gemini API key; after Google's June change, a personal Google AI Pro plan no longer covers it. Grok runs on your xAI key. Open models can run free on your own machine.
• It does not resell or meter model usage. There are no credits. Universe charges for the workspace software; the model bill stays between you and the provider, and your plan's own limits still apply.
• You can see who paid. Hover any agent answer to see which account ran it and what it cost.
• Several accounts, in order. Add a plan and an API key, and when the first is rate-limited the next runs. The session says so.

One caution in the other direction. A Universe workspace owner can share a login with the team: members use their own logins (the default), a shared account, or their choice, with an optional monthly cap per member. Codex and Gemini can pool only an API key, not a subscription login. Under the consumer terms quoted below, a personal Claude Pro or Max login is not yours to share, so put a company API key in the shared slot.

If you live in the terminal and use one model, you do not need any app for this: Claude Code or Codex on its own is the most direct route to your plan. Universe is for when you want agents working across your browser, files and connected apps, with several engines side by side. The full list is on features, plans are on pricing, and switching engines mid-task is covered in running Claude Code, Codex, Gemini and Grok from one app.

Red flags: when a tool is putting your account at risk#

• It asks for your session cookie or a token copied out of a browser's developer tools. No provider's sign-in works that way. Anthropic's page says sign-in "must complete through Anthropic's own flow."
• It asks you to paste auth.json or .credentials.json into a web form. OpenAI says to treat that file like a password.
• It sells "shared" Claude, ChatGPT or Gemini access at a fraction of the price. Anthropic's Consumer Terms say "You may not share your Account login information, Anthropic API key, or Account credentials with anyone else. You also may not make your Account available to anyone else." OpenAI's say "You may not share your account credentials or make your account available to anyone else." You are buying access to somebody else's account, which can be closed at any time.
• It runs turns on your personal plan from its own servers, for many users. That is close to what Anthropic means by "on behalf of their users".
• It will not say which route it uses. Assume route 2 or 3.
• It shows no record of which account paid. If an API key is set in your environment, Claude Code uses it once you approve it, and you find out from an invoice. Run /status to see which credential is active.

The cost of getting this wrong is not a warning email. Anthropic and Google both say they may act without notice, and the account at risk is the one you use every day.

Checklist before you connect a subscription to any app#

1. Which route is it? Official CLI (1), its own subscription login (2), or proxy (3)? If the app will not say, assume 2 or 3.
2. Where do you type your credentials? On the provider's own page (claude.ai, chatgpt.com, accounts.google.com)? If an app asks you to paste a session cookie or a token copied from another tool, stop.
3. Is the CLI unmodified? For Claude, Anthropic's permission covers the unmodified Claude Code binary.
4. Does it run on your machine or on a server? A hosted service running turns on your personal plan for you is what "on behalf of their users" describes.
5. Can you see which account paid for each turn? If not, you may find out about an API key charge from your invoice.
6. Is an API key set in your environment? Check /status in Claude Code, and echo $ANTHROPIC_API_KEY.
7. Is anyone else using your login? Personal plans are for one person.
8. Did you re-read the provider pages this month? Anthropic has said it will share changes before they take effect. Watch its help center and the Claude Code legal page.

If you left OpenClaw over the Claude rules, OpenClaw alternatives for Mac compares the options. If you use Codex with tools of your own and hit a refusal, see fixing "MCP tool call requires approval". For a definition of the category these apps belong to, see what an AI agent workspace is. To try Universe with the accounts you already have, download it; it runs on Apple silicon or Intel Macs with macOS 13 or later and is free to start.