Eligible ChatGPT Pro and Business subscribers can claim OpenAI API credits and use them in applications through Vercel AI Gateway. Link your subscription to an OpenAI API organization, then connect a key from that organization using Bring Your Own Key (BYOK). The credits stay in your OpenAI account.
This gives you a way to fund an application's first API requests with a benefit from a subscription you already have. For example, you could use the included credits to add a support-ticket summarizer to an internal dashboard.
Copy link to headingWhich ChatGPT plans include API credits?
Eligible paid Pro and Business plans include API credits, with amounts based on the subscription or paid seat mix.
Business credits are pooled, with a US$500 monthly workspace cap. Plus, Free, and trial subscriptions don't qualify. Your eligible paid subscription must have been active for at least seven days before claiming.
The first grant is prorated for the remaining billing period. Annually prepaid Business workspaces receive their annual grant upfront, capped at US$6,000, with the first grant prorated for the remaining term. Check the amount and expiration displayed during activation.
Copy link to headingHow do the credits work with AI Gateway?
OpenAI applies the grant to usage billed to the linked API organization. Gateway uses that organization's API key to authenticate requests to OpenAI, so qualifying usage can draw from its credits. Your application authenticates separately to Gateway with an AI Gateway key.
AI Gateway BYOK has no added markup or fee, but it requires purchased AI Gateway credits. If your OpenAI credentials fail, Gateway can retry with its system credentials and charge the Gateway balance. Plan for that possibility before sending requests.
Copy link to headingWhy use AI Gateway instead of calling OpenAI directly?
AI Gateway is useful when you want shared request logs and routing controls across the models your app uses. You can start with OpenAI and add another provider later while keeping those operations in one place. Your app can call Gateway from any cloud or local environment; hosting on Vercel isn't required.
For the support-ticket example, Gateway request logs let you inspect latency and token usage alongside the provider and billing route. If you later evaluate another model, its requests appear in the same view. Gateway also handles configured provider and model fallbacks, so your application doesn't have to implement every retry route itself.
You can keep the OpenAI SDK for supported Responses API or Chat Completions calls. Configure the Gateway base URL, authentication, and model identifier, then connect your OpenAI key through BYOK. The AI SDK example below is another client option.
Copy link to headingClaim the benefit and connect your app
Copy link to heading1. Link the OpenAI organization that will fund the app
Use the same email address for ChatGPT and the OpenAI API account. You must own both the eligible subscription or Business workspace and the receiving API organization. A Business admin role alone isn't sufficient.
In ChatGPT, open Settings > Billing for Pro, or the workspace billing settings for Business. Start the API credits claim flow, select the intended organization, and confirm. You can also claim from API Platform billing.
Open the linked organization's API billing page and confirm that the grant appears. A subscription can fund one linked organization, so choose the organization where the app's project will live.
Copy link to heading2. Add that organization's key to AI Gateway
Create a project API key in the credited OpenAI organization. In your Vercel team's AI Gateway settings, open Bring Your Own Key (BYOK), choose OpenAI, and select Add Key. Enter the provider key, leave it enabled, and use Test API Key to validate it.
Check which key is marked In Use if the team already has OpenAI credentials. Adding another key doesn't automatically replace the active one. The selected key needs to belong to the organization receiving the grant.
If your team enforces Zero Data Retention (ZDR), Gateway skips BYOK keys unless they're marked ZDR-compliant. Only mark a key compliant if your own agreement with OpenAI supports that designation; review the BYOK requirements before proceeding with ZDR enabled.
Copy link to heading3. Send a request through Gateway
Use Node.js 22.18 or later and install the AI SDK with pnpm add ai. Create an AI Gateway API key for the same Vercel team and set it as AI_GATEWAY_API_KEY in your server or local shell environment. Keep both keys out of browser code and source control. The text generation setup covers Gateway authentication and installation.
Save this example as summarize-ticket.mjs and run node summarize-ticket.mjs:
import { generateText } from 'ai';
const { text } = await generateText({ model: 'openai/gpt-6-astra', prompt: `Summarize this fictional support ticket for an agent.Include the problem and the next question to ask.Ticket: The customer can sign in but cannot download invoicesfor September. Downloads for August still work.`, providerOptions: { gateway: { only: ['openai'], }, },});
console.log(text);The script prints a generated summary. The provider filter keeps the request on OpenAI, where your grant applies. It doesn't disable the fallback from your key to Gateway's system credentials.
Copy link to heading4. Check which account paid for the request
Open the request in AI Gateway Logs and inspect its billing and routing details. Confirm that it used BYOK, then check usage and the grant balance in the linked OpenAI organization. A successful response alone doesn't establish that your promotional credits funded it.
Copy link to headingCan you use Codex with AI Gateway?
You can connect OpenAI Codex to AI Gateway to route coding requests through Gateway and monitor their usage alongside your application's traffic. OpenAI excludes Codex usage through API keys from the ChatGPT API credit benefit, so plan separate funding for this workflow.
With the Codex CLI installed, run npx vercel@latest ai-gateway setup --agent codex. The setup command configures a Gateway model provider in ~/.codex/config.toml and makes an AI Gateway key available to Codex. Add --dry-run to preview the configuration changes; the command backs up files it changes.
After setup, start codex and confirm that its requests appear in your AI Gateway Overview. You can use this connection for development while reserving the claimed API grant for eligible requests from the application you're building.
Copy link to headingWhat happens when the credits expire or run out?
OpenAI spends the grant before purchased API credits. After the grant is exhausted, requests can continue against an available purchased balance. With no credits left, OpenAI rejects requests, and Gateway's credential fallback may continue them against your Gateway balance.
Unused grants expire at the end of the applicable billing cycle. Renewals depend on continued eligibility and the benefit remaining available. Keep track of the provider balance when deciding how much ongoing API usage your app can support.
Gateway budgets exclude spend through your own credentials. Review provider-side usage controls as well as your Gateway budget. The grant also leaves model and rate limits in place, and it doesn't pay for hosting the application.
Copy link to headingWhat should you check if your ChatGPT API credits aren't being used?
If the grant is missing, the amount looks wrong, or you need help with eligibility, organization linking, or expiration, contact OpenAI Support. OpenAI manages the credit benefit and can review your account. Gateway's request logs show how a request was routed; they don't confirm how OpenAI applied a promotional grant.
For the Gateway connection, check:
In Bring Your Own Key (BYOK), confirm that the OpenAI key is enabled and marked In Use. Use the key test to check the credentials, and confirm the key belongs to the organization you linked when claiming.
Open the affected request in AI Gateway Logs and inspect the provider and credential route. If it used system credentials, that request used Gateway billing. Review the routing attempts for an error from the BYOK attempt.
A successful key test doesn't establish credit eligibility or availability. If the request used your OpenAI key but the credit balance still looks unexpected, take the account and request details to OpenAI Support for investigation.
Copy link to headingFrequently asked questions
Copy link to headingDoes ChatGPT Plus include these API credits?
No. This API credit benefit applies to eligible paid ChatGPT Pro and Business subscriptions. A Plus subscription does not qualify for the grant.
Copy link to headingAm I spending my ChatGPT usage allowance when my app calls the API?
Your app draws on the linked OpenAI API organization's balance. Claiming the benefit gives that organization a separate API grant; requests do not consume your ordinary ChatGPT message allowance.
Copy link to headingCan I use an OpenAI key I already added to Gateway?
Yes, if the active key belongs to the API organization that received the grant and has the permissions your request needs. A key from another organization won't draw on those credits, even if you own both organizations.
Copy link to headingCan these credits pay for Codex through an API key?
No. OpenAI excludes Codex usage through API keys from this benefit. The grant also doesn't increase signed-in Codex usage limits.
Copy link to headingWill my app stop when the ChatGPT API grant runs out?
Not necessarily. OpenAI can continue using purchased API credits, and a failed BYOK request can trigger Gateway's system-credential fallback. Review both balances and billing settings instead of treating the grant amount as a hard spending cap.
Copy link to headingCan I keep using the OpenAI SDK with AI Gateway?
Yes. AI Gateway supports OpenAI-compatible Responses and Chat Completions calls from the OpenAI SDK. Configure the Gateway endpoint, authentication, and model identifier, and check support for the features your application uses.