Sign in with ChatGPT explained: using your Plus or Pro plan in Devin, Notion, Vercel and other apps

The Sign in with ChatGPT button started as a simple login option. On September 29, 2026, at DevDay, OpenAI made it something bigger: Plus and Pro subscribers can now let third-party tools spend the AI usage included in their ChatGPT plan. At launch that covered 16 partners, from Cognition's Devin coding agent to Notion, Vercel and open-source agents such as OpenClaw and OpenCode. This explainer separates the two things the feature does, identity and plan usage. It covers what each one shares and what limits you control, then explains how developers integrate it. It is based on OpenAI's help center, product and developer documentation and terms, checked on October 12, 2026.

The short version

• Two separate permissions. Identity sign-in logs you in and shares your name, email address and profile picture. Plan usage lets an approved app run AI requests against your ChatGPT plan. You can grant the first without the second.
• Who can use plan usage: ChatGPT Plus and Pro subscribers, per OpenAI's product docs and DevDay recap. One help article also lists Go (see below).
• Where it comes from: the Work and Codex usage already included in your plan. It isn't a new allowance.
• Your controls: a weekly cap for each app in Settings > Usage, an optional switch to let apps use purchased credits, and Disconnect under Settings > Security and login.
• Developers: plan usage is open to open-source and local apps and selected private clients. Commercial apps need approval and go on a waitlist.

What is Sign in with ChatGPT?

OpenAI's help center describes it as "an identity-provider sign-in option that lets you use identity information from your ChatGPT account," much like "Sign in with Google." According to The New Stack, it first launched in July as an identity-only beta. Identity sign-in is "available globally to authenticated ChatGPT users, including users in Enterprise organizations." Its first partners included Airtable, GitLab, HubSpot, Notion, Supabase and Vercel.

The DevDay change added plan usage. OpenAI's recap says the feature now lets you "spend [your] ChatGPT plan allowance across 16 partners," that you can control how much each partner uses, and that "identity sign-in is available globally. Plan usage is available to Plus and Pro users in participating tools."

Which apps support it?

OpenAI's product documentation sorts partners into three groups.

Last verified: October 12, 2026

The 12 commercial apps and 4 open-source integrations make up the 16 launch partners. An OpenAI spokesperson told The New Stack: "We're starting with these 16 launch partners today, and we'll expand quickly."

Plan usage doesn't always cover everything an app does. The New Stack reports that in Dactyl, only the agent's code-writing step draws on your ChatGPT plan. Compiling, image generation, 3D assets and publishing still use Dactyl's own credits. Apps "may charge their own fees," OpenAI says, and may limit the option to some of their own plans.

Identity vs plan usage: what does each share?

Identity sign-in gives the app your name, email address and profile picture. OpenAI's help center says it doesn't share your ChatGPT conversations or memory, your files or tokens, or your billing information. Any other access an app asks for needs a separate permission, so review the request before you approve it.

Plan usage is a second permission, shown as "Use your ChatGPT plan." If you approve it, the app can send AI requests that count against your plan. OpenAI says this does not give the app your conversations or memories, and doesn't share an OpenAI API key.

How usage counts, per OpenAI's documentation:

• Requests draw on "the ChatGPT Work and Codex usage included in your plan." Connecting an app adds no new allowance.
• "Usage rates may differ between an app and ChatGPT," and OpenAI doesn't publish per-request rates for partner apps.
• When you hit your overall plan limit, the app stops using your plan unless you've allowed it to use credits and have credits available. An app that can't use your plan "won't automatically switch to another paid option."

How do you set limits or disconnect an app?

• Per-app caps: open Settings > Usage, find App limits and choose Manage limits. Each cap is weekly and set as a percentage of your overall weekly plan usage. OpenAI stresses that it's "a cap, not a separate pool of usage," so an app can hit its cap while your plan still has usage left. Signing in again doesn't restore usage.
• Credits: off by default. If your account offers the option, one setting lets apps you've allowed keep going on purchased credits after your plan limit. An app's limit must be set to 100% for this to work. Automatic credit purchases are a separate setting and can lead to charges.
• Disconnect: go to Settings > Security and login > Login connections > Sign in with ChatGPT, choose the app, then Manage connection > Disconnect. Signing out inside the app doesn't disconnect it, and disconnecting doesn't reverse usage already recorded.

Can Go or Business users use their plan in other apps?

This is where OpenAI's own pages disagree. The DevDay recap, the September 29 release note, the product docs and the developer docs all name Plus and Pro. The help article "Using your ChatGPT plan in other apps and sites" says plan usage in commercial apps "requires a Go, Plus, or Pro plan." Until OpenAI reconciles them, treat Go support as unconfirmed. No page mentions Business plan usage.

For workplaces, identity sign-in is enabled by default for organizations without an explicit identity policy. Existing allow and deny lists still apply, and global admins manage application permissions under External access in the OpenAI Admin Console.

How do developers integrate Sign in with ChatGPT?

OpenAI's developer docs at developers.openai.com/siwc offer three integration types: on your website, in your ChatGPT plugin, and in your open-source app. Access is gated:

• Identity sign-in for commercial apps is "currently available to selected commercial partners through a limited trial."
• Plan usage is open to open-source projects, personal projects that run locally and selected private apps. Paid or remotely hosted apps must join a waitlist first, and OpenAI has an interest form.

The flow, as OpenAI's quickstart describes it:

1. Your app starts an OpenID Connect sign-in with PKCE.
2. The user authenticates and consents with OpenAI.
3. OpenAI returns an authorization code to your registered callback.
4. Your backend exchanges the code and verifies the ID token.
5. Your app links or creates a local account and issues its own session.

Identity uses the openid profile email scopes. Plan usage needs extra scopes the user must approve. OpenAI's open-source guide names offline_access, resource.invoke and chatgpt.tokens.use.direct, with https://api.openai.com/v1 as the resource. Once those are granted, your app calls the public Responses API (POST /v1/responses) with the user's OAuth access token instead of an API key. OpenAI says not to point it at ChatGPT's internal backend-api endpoints.

The plan-usage path is narrower than the regular API. OpenAI's preview-limitations page says requests must set store: false and stream: true and send the full history in input. It also says many parameters, including temperature, top_p and previous_response_id, must be omitted. Image generation, file search, Code Interpreter, native computer use and hosted MCP connectors aren't supported. Use GET /v1/models to see which models the signed-in account can use. When a user hits a cap, requests fail with errors such as subscription_sharing_usage_limit_exceeded.

OpenAI's Sign in with ChatGPT terms (page dated September 29, 2026) prohibit "pooling, transferring, reselling, gifting, or sharing ChatGPT plan usage." They also bar using one user's plan to serve another and providing "general-purpose API access for other tools or unrelated requests." The button must read "Continue with ChatGPT," with comparable prominence to your other sign-in options.

Why it matters

For tool makers, The New Stack's coverage puts the economics plainly: users already pay for ChatGPT, so developers no longer have to cover every user's token costs to onboard them. For subscribers, a $20 Plus or $100–$500 Pro plan now reaches beyond OpenAI's own apps. That makes the weekly caps worth setting, because a busy coding agent can use up the same allowance you need in Codex and Work. For how ChatGPT's plans and allowances compare, see our ChatGPT plans comparison. For the developer side of OpenAI's agent platform, see the Agents API public beta, and for the Codex-vs-competitors picture, Claude Code vs Codex vs Cursor.

What's confirmed, reported and uncertain

• Confirmed (OpenAI):
  • the September 29 expansion and the partner list and groups
  • what identity sign-in shares
  • that plan usage draws on Work and Codex usage
  • weekly per-app caps, the credit opt-in and disconnect steps
  • the OAuth/OIDC flow, scopes and Responses API restrictions
  • the terms' anti-resale rules
• Reported: the July 2026 identity beta, the spokesperson quote and Dactyl's partial coverage (The New Stack).
• Uncertain:
  • whether Go subscribers can use plan usage (OpenAI's pages conflict)
  • per-request usage rates in partner apps
  • when commercial apps outside the trial can join
  • when Lovable support arrives