Vercel Blog→ original

Vercel launches BYOC on AWS and an agent authorization beta as AI token usage quadruples

Vercel has expanded its Enterprise Apps and Agents platform, which since June has served as a single control center for developers, apps, agents, and tokens. Bring Your Own Cloud on AWS is now generally available, so compute and data stay inside the customer's own cloud. Enterprise-Managed Authorization in Vercel Connect has entered public beta: agents access services on an employee's behalf through the corporate IdP. Enterprise customers' token spend on AI Gateway grew 4x in three months. At Klaviyo, 512 employees shipped 356 apps in two weeks. • BYOC on AWS: GA • Enterprise-Managed Authorization: public beta • Deploying an internal app at Klaviyo: from two weeks to three minutes

AI-processed from Vercel Blog; edited by Hamidun News
Vercel launches BYOC on AWS and an agent authorization beta as AI token usage quadruples
Source: Vercel Blog. Collage: Hamidun News.
◐ Listen to article

Vercel has expanded its Enterprise Apps and Agents platform: Bring Your Own Cloud (BYOC) on AWS is now generally available, and Enterprise-Managed Authorization in Vercel Connect has launched in public beta. The trigger is demand. According to Vercel, enterprise customers' token spend on AI Gateway grew 4x over the past three months. OpenAI reported that from August 2025 to June 2026, the number of Codex users who are not developers grew 189-fold.

Everyone builds, but IT is accountable

Vercel calls these users citizen developers. A marketer builds a landing page in ChatGPT, an analyst runs queries and builds a report in Claude Code, an engineer works in Codex, OpenCode, or fx. Each knows their own part of the business better than the IT department does, and can now write the software they need themselves. The flip side is that employees use AI tools whether or not those tools are approved. If a vibe coder builds a login page incorrectly, private data can end up on the public internet.

Enterprise Apps and Agents has been available to customers since June. It is a single destination for deployments, for both engineers and non-technical staff. Policies for access, providers, models, data storage, token budgets, and integrations are set centrally through the company's existing IdP.

Klaviyo offers a telling example. Shipping a simple internal app used to take two weeks: chasing IT, filing requests with engineers, nudging the platform team. The company built an internal platform on Enterprise Apps and Agents, and now an idea can be deployed in three minutes. In the first two weeks, 512 employees shipped 356 apps. Klaviyo's senior AI architect Mohamed Ali says Vercel let the company build guardrails instead of walls.

Four domains of control

Vercel splits governance into four areas and ties a set of tools to each:

  • Deployment: Enterprise Managed Users and RBAC.
  • Access: Vercel Passport, Deployment Protection, and Vercel Firewall.
  • Integrations and data: Vercel Connect and Enterprise-Managed Authorization.
  • Models and tokens: AI Gateway, including US-only inference and budget limits.

Observability and security monitoring live in one place: analytics, audit logs, and a Security Dashboard across all projects. Secure Compute provides private connectivity to corporate systems. BYOC on AWS runs application functions in the customer's own account, under its permissions and network restrictions. The company gets the full Vercel developer experience, while compute and data never leave the AWS perimeter.

Agent access to systems is a separate problem. The usual approach is to set up long-lived environment variables with broad permissions to CRMs, databases, and Slack. Vercel Connect instead creates connectors with limited access and issues a token for each agent task, so the agent sees only the context it needs.

The new Enterprise-Managed Authorization adds convenience: when an employee signs in to Vercel through the IdP, agents use connected services without a separate authorization prompt. Connect works together with Passport, so an individual's permissions also apply at the integration level. If they cannot see data in the source system, they will not get it through the app either.

By default, every deployment can be made private, and access to internal apps can be tied to IdP sign-in.

Where to start

For IT teams, the practical order is as follows. First, take inventory: which AI tools employees already use and which long-lived keys sit in agents' environment variables. Next, set up a single deployment point with privacy by default and IdP sign-in. After that, add token budgets and model rules, such as US-only inference if compliance requires it.

BYOC on AWS suits those who need to keep data in their own cloud: it is already generally available. Enterprise-Managed Authorization is still in beta, so it makes sense to try it first on non-critical integrations. The 189-fold growth in non-developer Codex users shows that the question of whether to let everyone build has already been settled by the employees themselves. What remains is deciding where all of this will live and who is responsible for it.

ZK
Hamidun News
AI news without noise. Daily editorial selection from 50+ sources. A product by Zhemal Khamidun, Head of AI at Alpina Digital.
✦ Autonomous AI Agents

Delegate Deep Research, Coding & Workflows to an AI Agent

Deploy autonomous multi-step execution with NomadNet Agent (agent.nomadnet.ai) — automated web research, data synthesis, code generation, and complex task orchestration.

Launch NomadNet Agent →NomadNet AI Platform
What do you think?
Loading comments…