What happened
Vercel says Token Pass credit now applies automatically to included-model usage through AI Gateway, regardless of whether a request authenticates with an API key, an OIDC token, or an fx login. Previously, a subscriber needed a dedicated Token Pass API key for that access; the changelog says this key is no longer required. The update is available now to Token Pass subscribers on Pro teams. Vercel does not give a release date beyond “now” in the supplied announcement.
A dedicated Token Pass key remains an option. Vercel says subscribers can create one from the dashboard or CLI, scoped only to featured models and providers. This preserves a way to make credentials specific to Token Pass rather than using a standard key across AI Gateway. The source does not describe any change to the price of Token Pass, the amount of credit, the included models, or the billing period; it announces a change to how credits apply and how requests behave after the balance is used.
The important distinction appears when the Token Pass balance is exhausted for the billing period. Requests authenticated with a standard API key or OIDC token fall back to the team’s AI Gateway credits. Requests made with a Token Pass key stop until the next billing period. The source does not specify a fallback for fx logins, so teams should not assume the stated API-key/OIDC behavior applies to fx as well. Vercel specifically recommends that subscribers who created a Token Pass key during setup switch to a standard API key or OIDC token if they want fallback.
The announcement also clarifies data-retention behavior. Requests served through the Token Pass provider remain Zero Data Retention (ZDR). Fallback requests follow the team’s setting. If team-wide ZDR is enabled, Vercel says fallback requests route only to ZDR providers. That means the switch from pass credit to team credit does not automatically imply the same provider handling in every team: the team’s ZDR configuration governs the fallback path.
What the archive adds
Earlier coverage described Token Pass as a prepaid team-level budget that bundles access to multiple models. Vercel Token Pass bundles GLM and Kimi credits for Pro teams established that framing; DeepSeek V4.1 Flash 1017 is now included in Vercel Token Pass covered an included model. This is a new operational fact about credential handling and exhaustion behavior, not a new Token Pass launch.
Trend signal
No supplied SITE_PATTERNS entry directly covers subscription-credit routing, credential choice, or post-exhaustion fallback. The change is a useful product detail for Token Pass users, but treating it as evidence for a broader agent-economics or protocol trend would stretch the supplied material.
Why now
Vercel presents the change as available now, but the announcement gives no rationale or timing context. It does not connect the update to a pricing change, a new model addition, or a broader product launch. The practical timing is therefore tied to subscribers’ current setup: teams using a dedicated Token Pass key should review it if they expect requests to continue on team credits after pass credit is spent.
Why it matters
This is a small but consequential choice in the request path. A team can now use its usual AI Gateway credentials for included Token Pass models rather than maintaining a separate key just to access the subscription credit. That can make an existing integration simpler, but the more important operational difference is at the limit: standard keys and OIDC tokens can continue on team credits, while Token Pass keys stop.
For an application that must remain available, that distinction should be deliberate. A team may prefer stopping rather than incurring team-credit usage, or it may prefer continuity and accept the fallback. The update makes those outcomes legible enough to design around, though it does not disclose pricing or cost controls. ZDR adds another configuration dependency: fallback requests follow team settings, and team-wide ZDR restricts routing to ZDR providers. Credential selection, budget expectations, and data-retention policy now need to be considered together, not as isolated setup details.
Limits and uncertainty
The announcement establishes availability for Token Pass subscribers on Pro teams, eligible authentication methods, and the stated fallback behavior for standard API keys and OIDC tokens. It does not establish that fx logins receive fallback, nor does it quantify balances, prices, included-model limits, or what happens after team AI Gateway credits are depleted. It also does not say whether existing deployments are migrated automatically or whether any code changes are needed. The ZDR statement is limited to requests served through the Token Pass provider and the team’s fallback configuration; it does not provide model-by-model retention details. Teams should confirm their credential type and settings in their own environment rather than infer behavior beyond the stated cases.
Skill to build now
For a service using Token Pass, test the exhaustion path in a non-production environment this week. Record which credential the application sends, verify whether a standard API key or OIDC request falls back to team AI Gateway credits, and inspect whether team-wide ZDR restricts the fallback provider set as expected. If the service currently uses a dedicated Token Pass key, make an explicit decision: keep requests stopping at the period limit, or switch to a standard key/OIDC token for the documented fallback behavior. Document that choice with the team’s budget and data-handling assumptions.
Reader takeaway
Token Pass now works through ordinary AI Gateway credentials, but credential type determines what happens when its credits run out. Standard API keys and OIDC tokens fall back to team credits; dedicated Token Pass keys stop. Check that behavior alongside your team’s ZDR settings before relying on continuity.