When everyone in the company builds internal tools with AI, the hard part is not shipping — it is the four questions that follow an agent into use, which Vercel summarizes from a year of internal experience: who is allowed to use each agent? How do we keep internal agents internal? Which data and systems are agents allowed to touch? Which models are they using, and how much do they cost? Over the past year Vercel employees shipped hundreds of agents and internal apps on eve and the Agent Stack; getting to production was the easy part, and governance was the real work. Enterprise Apps and Agents productizes those lessons: ownership, access, and security become defaults your builders inherit, not projects your platform team queues.
| Platform component | Security implementation |
|---|---|
| Vercel Passport | Puts every internal app and agent behind your identity provider by default |
| Vercel Connect | Gives agents short-lived, scoped credentials for Slack, GitHub, Snowflake, Salesforce, and Linear |
| Enterprise Managed Users | Full lifecycle control over every Vercel and v0 user through your existing directory |
| Bring your own cloud on AWS | Runs apps and agents inside your own AWS account (Private Beta) |
Passport: Internal Means Internal, by Default
Internal tools are most companies’ first agents, and “internal” used to be a per-project setting — one employee forgetting to make one deployment private risked exposing sensitive systems and data. Vercel Passport inverts the assumption: private is the state that exists the moment a deployment does, access authenticates against employee identity, every entry is auditable, and admins set policy centrally. Connect your IdP once (Okta, Microsoft Entra, Auth0, or any OpenID Connect-compatible provider) and Passport applies it across every deployment automatically. Currently in Beta.
Connect: Short-Lived Tokens Instead of Long-Lived Keys
Passport governs who can reach an agent; the agent itself still needs access to data and systems — which is exactly what makes agents useful and dangerous. The traditional pattern parks long-lived credentials in environment variables, provisioned for everything the agent might ever do. Vercel Connect consolidates OAuth, OIDC, and secret injection into one product that replaces static keys: the agent stores no secret and instead requests short-lived credentials as it works — tokens granted per task, expiring when the task completes. Supported today: Slack, GitHub, Snowflake, Salesforce, and Linear, plus anything reachable over OAuth or an API. Currently in Beta.
Enterprise Managed Users: The Answer to Account Sprawl
When everyone is a builder, account sprawl is a silent failure mode: seats provisioned by no one, access lingering after role changes or departures, no single record of who did what. Managed Users builds on SAML SSO and Directory Sync — a seat exists the moment the directory says it should, and off-boarding removes access the moment the directory does. Group-based access controls, deployment protection, and MFA enforcement apply org-wide, and every action lands in a single audit trail; your existing Okta or other SAML/OIDC directory can now govern v0 as well. Currently in Private Beta. Alongside it, v0 now connects to Snowflake: anyone can build data apps directly on the warehouse, access still flows through your IdP, and data stays inside the boundary.
BYOC on AWS: Workloads in Your Own Account
For large enterprises the boundary question goes past private deployments — the workloads themselves must run on infrastructure the security team owns and audits. Bring your own cloud (BYOC) runs your compute, build artifacts, and data inside your own AWS account and VPC, with Vercel operating the control plane on top; apps and agents reach private backends the same way anything else in the account does, and source code never leaves your CI. Engineers keep the Vercel experience; the security team keeps its network controls, audit evidence, and account. Currently in Private Beta on AWS.
What This Means for Platform Teams: Make the Safe Path the Default
Vercel’s own summary of the difference: traditionally, ideas died in security review for good reason — breach risk outweighed innovation gain. Build the controls into the platform itself and three things follow: secure prototyping at scale (anyone experiments on v0’s governed rails, and experimentation stops being a security exception); domain experts building their own tools (the people closest to a problem solve it instead of filing a ticket and waiting a quarter); and an immediate graduation path to production (a proven prototype ships on the platform it was prototyped on — no rebuild, no ban). One caveat for evaluators: Passport and Connect are in Beta while Managed Users and BYOC are in Private Beta — the governance blueprint is complete, but availability needs checking per component.
Sources
AI-assisted summary compiled from the sources above, reviewed by a human before publishing.