The catch is that every partner decides which of its own plans qualify, which models are covered, what still uses its own credits, and what happens when your ChatGPT allowance runs out.
Quick answer: A supported third-party app can use eligible requests against the ChatGPT Work and Codex allowance included with Plus or Pro, without receiving your OpenAI API key. The app gets your name, email address and profile picture through the normal sign-in flow, not your ChatGPT conversations or memory. Partner subscriptions, infrastructure costs, unsupported features and extra usage can still be charged separately.
The easiest way to understand the system is to split it into four layers.
Identity, permission, usage and billing are separate
| Layer | What actually happens |
|---|---|
| Identity | ChatGPT confirms who you are and shares your name, email address and profile picture. |
| Permission | You separately approve Use your ChatGPT plan. Signing in alone does not grant plan usage. |
| Usage | Eligible requests draw against the Work and Codex allowance already included with your plan. |
| Billing | Your ChatGPT subscription, ChatGPT credits and the outside app's own pricing remain separate. |
This explains why the same Continue with ChatGPT button can mean very different things in two products.
Products including Notion, Devin, Warp, Kilo Code and Vercel's v0 can also use a connected ChatGPT plan, subject to their own implementation and billing rules.
Seeing the login option does not automatically mean your ChatGPT subscription will pay for AI usage in that product.
Which ChatGPT plans can use it?
For commercial partner apps, OpenAI currently documents ChatGPT plan usage for Plus and Pro.
| ChatGPT plan | Sign in with ChatGPT | Use your plan in commercial partner apps |
|---|---|---|
| Free | Yes | No |
| Go | Yes | No |
| Plus | Yes | Yes |
| Pro | Yes | Yes |
| Business | Yes, subject to organization controls | Not listed as eligible |
| Enterprise | Yes, subject to organization controls | Not listed as eligible |
| Edu | Not specifically documented | Not listed as eligible |
OpenAI says Sign in with ChatGPT itself is globally available to authenticated users, with administrators able to control external app access in managed organizations.
Commercial plan sharing is narrower.
OpenAI currently describes Plus and Pro as the eligible plans.
There is a separate open-source route for tools such as OpenCode, Pi, OpenClaw and T3. OpenAI says users can connect ChatGPT accounts to supported open-source tools, but the documentation does not make Free and Go usage entitlement clear enough to treat that route as equivalent to commercial partner access.
What does your ChatGPT subscription actually cover?
Connecting another app does not create a second allowance.
Eligible requests use the same underlying capacity that powers Work and Codex. OpenAI also describes that pool as shared with products such as ChatGPT for Excel and Workspace Agents where they are available on the account.
That has an important practical consequence.
Usage inside another app can reduce what remains available elsewhere.
OpenAI lets users set a weekly app limit under Settings > Usage, expressed as a percentage of overall weekly usage.
That percentage is a ceiling. It does not reserve capacity or add more usage to your plan.
An outside app can therefore hit its own cap while you still have allowance remaining elsewhere, or it can consume enough of the shared pool to affect Work or Codex later in the week.
Usage rates can also vary between applications and operating modes. OpenAI warns that faster modes can consume plan allowance or credits more quickly.
Kilo's documentation additionally describes a five-hour usage window alongside the weekly allowance for supported Codex use.
The practical calculation is straightforward: connecting another app gives you another place to spend the allowance you already bought.
For someone paying for ChatGPT and also purchasing OpenAI-backed credits inside another tool, that can remove genuine duplication.
For someone already pushing Work or Codex close to its limits, the value is less obvious.
What does Sign in with ChatGPT still not pay for?
Three separate cost layers can remain.
Your ChatGPT subscription
You continue paying OpenAI for ChatGPT Plus or Pro.
That subscription provides the eligible allowance being used inside the partner product.
The partner's own software
The outside company can still charge for its product, infrastructure, services or premium features.
OpenAI's current Help documentation specifically says participating apps may make plan usage available only on certain app plans and that partner subscriptions and charges remain separate from your ChatGPT plan.
That already happens in practice.
Notion requires the user to belong to a Business or Enterprise workspace.
Devin makes the feature available on its self-serve Pro, Max and Teams plans, rather than its Free plan.
OpenAI's public Sign in with ChatGPT terms include a no-charge requirement for the self-serve integration route, but those terms also contain conditions around locally controlled runtimes and token storage. OpenAI's developer documentation sends paid and remotely hosted apps through a separate commercial process.
The commercial agreements for those partners are not public.
For a normal user, the partner's own billing page is therefore the safer guide to what the integration will actually cost.
Features outside the ChatGPT-plan route
Unsupported models and features can continue to use the partner's own meter.
That is one reason connecting ChatGPT rarely turns the rest of the product free.
OpenAI's self-serve preview route, for example, excludes several hosted capabilities including image generation, File Search, Code Interpreter, native computer use, hosted MCP/connectors and audio input.
Commercial partners can have different arrangements, but the restriction illustrates the wider point: sharing a ChatGPT plan does not automatically include every OpenAI-powered feature inside another product.
How Notion, Devin, v0 and Kilo handle it
The partner differences are the real story.
| App | Requirement on the app side | What your ChatGPT plan covers | What still uses the app's meter | When ChatGPT usage runs out |
|---|---|---|---|---|
| Notion | Business or Enterprise workspace; members rather than guests | Sessions using a specific supported GPT model | Auto mode, non-GPT models and Custom Agents | Notion asks you to switch to workspace credits |
| Devin | Self-serve Pro, Max or Teams plan | Supported GPT models, excluding some fast and priority variants | Non-OpenAI models and other Devin usage | Devin automatically falls back to its own quota, then on-demand credits |
| v0 by Vercel | No additional app requirement stated in v0's documentation | Eligible responses using supported OpenAI models | Image generation, image editing, delegated subagents and supporting agent work | v0 pauses and asks whether to continue using v0 credits |
| Kilo Code | ChatGPT sign-in is offered without a separate Kilo subscription requirement | Supported OpenAI model requests through the Responses API | Cloud Agent compute, Kilo Deploy and other paid Kilo services | The request normally stops with a limit error |
Kilo documents another edge case. If ChatGPT authentication cannot be renewed before a request is sent, certain failures may allow the request to use a saved API key or Kilo credits instead.
Warp says the Work and Codex usage included in a ChatGPT plan can cover AI requests in Warp Terminal and Warp Agent CLI. Its public launch documentation does not spell out the same level of limit and fallback detail, so users should still check Warp's current billing controls before running long agent jobs.
This variation matters more than the logo wall.
The feature can replace one source of AI inference cost inside another product while leaving the rest of that product's economics intact.
What happens when your ChatGPT allowance runs out?
OpenAI's side is reasonably clear.
If an app reaches the percentage limit you set for it, or your overall eligible plan usage runs out, plan-backed usage stops.
Participating apps can continue onto available ChatGPT credits only when several conditions are met:
- credit use for connected apps has been enabled
- the app's usage limit is set to 100 percent
- eligible ChatGPT credits are available
Credit use for apps is off by default.
Setting an app to 100 percent does not turn credit use on by itself.
Automatic credit purchases are another separate setting again. If automatic purchases are enabled, continued usage can create additional OpenAI charges.
OpenAI also says its own system does not automatically move a failed plan-backed request onto another paid OpenAI option.
That does not mean every partner stops.
Devin documents an automatic fallback to its normal Devin quota and then on-demand credits when ChatGPT usage is unavailable.
v0 pauses and asks the user whether to switch to v0 credits.
Notion asks the user to continue with workspace credits.
Kilo normally stops the request.
If an unexpected charge matters to you, check the partner's fallback behavior before starting a long-running agent job.
Does Sign in with ChatGPT share your chats?
No.
OpenAI says the standard identity flow shares:
- your name
- your email address
- your profile picture, if you have one
It does not independently share your ChatGPT conversations, memory, files, tokens, billing information or other ChatGPT account data.
Approving plan usage also does not give the partner your OpenAI API key or general access to your ChatGPT account context.
That privacy boundary only covers the ChatGPT connection itself.
The app still sees what you give the app
If you type a prompt into Notion, upload code through Kilo or give another agent access to a repository, the partner processes the information required to perform that work.
Kilo, for example, says requests and their code context pass through the Kilo Gateway before reaching OpenAI.
The partner's privacy policy therefore still matters.
The partner's own data rules still apply
Notion says its own AI subprocessor agreements, including its no-training commitments, continue to apply when a request is funded through the user's ChatGPT plan.
It also says those sessions do not appear in ChatGPT history or become ChatGPT memory.
That is an important distinction. Paying for inference through ChatGPT does not move the entire session into ChatGPT.
Apps can request more access separately
Sign in with ChatGPT can be followed by additional delegated permissions.
OpenAI gives access to a ChatGPT Ads account as one example of a separate permission an app may request.
Managed organizations can also create another visibility layer.
Notion says workspace administrators can see connection and disconnection events in audit logs, although they cannot see the amount of ChatGPT usage consumed.
The security question is therefore wider than whether the app can see your old chats.
You need to know what this particular product can read, what it can do and what information you are deliberately sending through it.
For the wider permissions problem, see The Always-On AI Assistant Is Here. What Should You Actually Let It Do?.
Does ChatGPT Plus include API access now?
No.
OpenAI still bills API usage separately from ChatGPT subscriptions.
Paying for Plus does not give you a general OpenAI API balance that you can spend across your own applications.
Sign in with ChatGPT creates a narrower path.
A participating product receives OAuth authorization to make eligible requests against your ChatGPT plan for that connected application.
There is no API key handed to the user or partner through that process, and the plan cannot simply be turned into general-purpose API credit for unrelated applications.
So a developer building their own product against OpenAI's API still needs normal API billing.
The new exception applies to supported integrations.
For the broader shift from simple monthly plans toward allowances, credits and workload-based economics, see The Unlimited AI Subscription Is Dying.
Who should actually connect an app?
You already pay for ChatGPT and buy OpenAI-backed usage elsewhere
This is the strongest case.
If part of a second product's AI bill can move onto an allowance you already pay OpenAI for, Sign in with ChatGPT can reduce duplicated spend.
Check exactly which models and features qualify first.
You already use Work or Codex heavily
Be careful.
Every eligible third-party request competes with other products drawing on the same allowance.
Saving money in one tool is less impressive if it makes your main Codex workflow hit its limit on Thursday afternoon.
The partner requires a more expensive plan
Notion and Devin show why the app-side gate matters.
An integration may technically save model credits while requiring an upgrade to the partner's own product.
Compare the upgrade cost with the usage cost you are actually replacing.
Most of the product's cost comes from compute or infrastructure
The benefit may be modest.
ChatGPT plan usage can cover model inference while cloud compute, deployments, image generation and other services remain separately billed.
You are on Free, Go or a managed plan
Sign in with ChatGPT can still be useful as an identity option where your account or administrator allows it.
Commercial third-party plan usage is currently documented for Plus and Pro.
You need normal API access
Use the API.
This feature is not a substitute for standard developer billing.
Eight things to check before connecting an app
- 1Confirm the product appears under plan usage, rather than sign-in only, in OpenAI's current partner documentation.
- 2Check whether the partner requires one of its own paid plans.
- 3Confirm which models and features actually use the ChatGPT allowance.
- 4Find out what happens when the allowance ends: stop, ask or bill elsewhere.
- 5Set a sensible app limit under Settings > Usage before handing the app a large agent task.
- 6Check whether connected apps are allowed to use ChatGPT credits, then inspect automatic credit purchases separately.
- 7Read the partner's privacy policy for the prompts, files and code you will send through the product.
- 8Know how to disconnect the authorization when you are finished.
That final step is easy to miss.
Signing out of a third-party app does not disconnect its ChatGPT authorization.
OpenAI says the app remains connected until the user goes to:
Settings > Security and login > Login connections > Sign in with ChatGPT
and disconnects the application.
Disconnecting stops future authorized use. It does not reverse usage already recorded or delete information the third party previously received.
OpenAI also notes that some sign-in-only connections may not show the same Disconnect control.
Choosely's verdict
Sign in with ChatGPT is worth paying attention to because it can turn one AI subscription into useful buying power across several products.
The saving is real when a partner moves OpenAI model usage you already pay for onto your existing Plus or Pro allowance.
The product details decide whether it is actually a good deal.
Notion and Devin can require paid partner plans. v0 still charges for parts of the workflow. Kilo keeps compute and deployment outside the ChatGPT allowance. Limit behavior also ranges from a hard stop to automatic fallback billing.
Before connecting anything, answer two questions:
Which charge does this replace?
What happens when the shared allowance runs out?
If those answers are clear and the numbers work, use it.
If they are not, another login button has not simplified much.
FAQ
Can I use ChatGPT Plus in other apps?
Yes. Participating apps that support plan usage can let Plus and Pro users approve a separate Use your ChatGPT plan permission. Eligible requests then count against the Work and Codex usage included in the plan.
Some partner products have their own paid-plan requirements.
Does ChatGPT Plus include API access now?
No. General OpenAI API usage remains separately billed.
Sign in with ChatGPT allows approved applications to make eligible plan-backed requests through OAuth without receiving an API key.
Does Sign in with ChatGPT share my chats?
No. OpenAI says the identity flow shares your name, email address and profile picture.
It does not independently expose your ChatGPT conversations or memory.
The partner still processes whatever information you deliberately send through its product.
Can an app see my memory, projects or custom instructions?
OpenAI says plan sharing does not grant access to conversations, memory or other ChatGPT account context.
Projects and custom instructions are not individually named in the documentation, so Choosely would not make a narrower claim beyond that documented boundary.
Do conversations in another app appear in my ChatGPT history?
Not necessarily.
Notion specifically says sessions funded through a connected ChatGPT plan are not saved into ChatGPT history or memory.
Other products retain their own records according to their own policies.
Do I still have to pay for the other app?
Possibly.
Partners can still charge separately for their subscription, infrastructure, services and premium features.
Some also restrict ChatGPT plan usage to particular partner tiers.
Does using another app count against my ChatGPT limits?
Yes.
Eligible requests use the same shared Work and Codex allowance rather than receiving a separate pool.
The rate at which usage is consumed can also differ between apps.
What happens when the allowance runs out?
It depends on the partner.
OpenAI plan usage stops unless eligible ChatGPT credits have been enabled for the app.
v0 pauses, Notion asks whether to switch to workspace credits, Kilo normally stops with an error, and Devin can automatically fall back to its own quota and paid credits.
Can I disconnect an app later?
Yes.
For supported plan connections, go to Settings > Security and login > Login connections > Sign in with ChatGPT, select the application and disconnect it.
Signing out of the partner product does not disconnect the ChatGPT authorization.
Can ChatGPT Business, Enterprise or Edu use their plan in other apps?
OpenAI currently documents commercial third-party plan usage for Plus and Pro.
Business and Enterprise accounts can use Sign in with ChatGPT for identity where organizational controls allow it. Edu is not specifically documented in the current Sign in with ChatGPT guidance.
Keep your AI stack visible
Using one subscription across several products can reduce duplicated AI costs. It can also spread usage, permissions and billing across more places.
Choosely helps you keep the tools your work depends on visible, track what changes and decide whether each part of your stack still earns its place.
The Change Brief
Get the week’s AI changes in one clear read
Pricing moves, tool launches, free-tier changes and practical stack updates, filtered for people who actually use these tools.
Stay ahead of AI without following it all day. We’ll send you what matters each week.
Continue reading
Related reads
AI Tool Recommendations
ChatGPT Plus vs Claude Pro vs Google AI Pro: Which $20 Plan Earns Your Money?
ChatGPT Plus is the safest all-purpose AI subscription, Claude Pro suits focused production work, and Google AI Pro offers the strongest bundle.
AI Strategy
The Unlimited AI Subscription Is Dying
AI subscriptions are shifting toward credits, token metering and usage-based billing. Here is what Copilot, Codex and Claude reveal about the real cost of AI workflows.
AI Strategy
MCP Explained: How AI Agents Connect to Your Tools and What They Can Access
MCP is the protocol many AI agents use to connect to tools, files, apps, and workflows. Here is what it is, how it works, and what permissions to review before you enable it.
AI Strategy
The Always-On AI Assistant Is Here. What Should You Actually Let It Do?
Persistent AI assistants can remember, act and continue working after the prompt ends. The real product decision is how much authority they should receive.
