Vercel Makes Token Pass Credit Apply Automatically in AI Gateway.

Token Pass credit on Vercel no longer needs its own API key to work. As of the 6 November changelog entry, Token Pass credit applies automatically to included model usage through AI Gateway, whatever method a request uses to authenticate: API key, OIDC token, or fx login.

Before this change, you had to generate a dedicated Token Pass key and route requests through it to draw on that credit. Miss that step and you were billed from your team's regular AI Gateway credits instead, even if you had Token Pass sitting unused. Now the gateway checks for included-model usage first and applies Token Pass credit by default, regardless of which key or login made the call.

You can still create a Token Pass API key from the dashboard or the CLI if you want one scoped only to featured models and providers. That option has not gone away. It is just no longer the only way in.

When Token Pass credit is exhausted for the billing period, requests made through API keys, OIDC tokens, or fx logins fall back to the team's standard AI Gateway credits, so nothing breaks. A dedicated Token Pass key behaves differently: it stops working entirely until the next billing period starts. That is the one behavioural difference worth remembering if you have built automation around a Token Pass key specifically.

What is Token Pass on Vercel? Token Pass is a Vercel subscription that gives teams included credit for usage of featured models through AI Gateway. Previously that credit only applied when requests were made with a dedicated Token Pass API key; it now applies automatically no matter how the request authenticates.

Which Vercel plan includes this change? The automatic credit application is available now for Token Pass subscribers on Pro teams, according to Vercel. The changelog does not specify whether or when it reaches other plan tiers.

What happens if I still use a dedicated Token Pass API key? You can keep using one, and it still works for scoping access to featured models and providers only. The difference is that a dedicated key stops working once its credit runs out for the billing period, whereas requests through other auth methods fall back to the team's regular AI Gateway credits instead.