Vercel has introduced a set of enterprise-focused tools for building and governing AI applications and agents, including the open-source eve framework, AI SDK 7, AI Gateway, Passport, Connect, and bring-your-own-cloud deployment on AWS.
Vercel announced an enterprise platform for AI apps and agents that combines its new eve framework with governance, identity, deployment, and model-routing tools.
The company’s announcement, published on the Vercel blog, frames the effort around a common enterprise problem: organizations want to adopt AI agents while keeping access controls, security boundaries, data handling, and infrastructure choices under internal governance.
Vercel introduced eve as an open-source agent framework and described it as a production-oriented foundation for building AI agents. In a separate Vercel post titled “Introducing eve,” the company compares eve’s role to what Next.js provides for web applications: a structured framework intended to move projects from prototypes toward deployed software.
According to Vercel, eve includes durable execution, sandboxed compute, human approvals, evaluations, AI Gateway fallbacks, and a file-based structure for defining agents. The Register also reported that eve is released under the Apache-2.0 license, corroborating Vercel’s positioning of it as an open-source framework.
Those features matter for enterprises because many internal AI tools need to run longer tasks, call external tools, ask humans for approval before sensitive actions, and produce logs or evaluation data that teams can audit. Vercel’s materials present eve as part of a broader “Agent Stack,” rather than as a standalone demo framework.
Vercel says AI SDK 7 is the TypeScript layer that eve is built on. In its AI SDK 7 announcement, the company says the release adds cross-provider agent capabilities such as reasoning control, tool approvals, durability, sandbox support, telemetry, realtime voice, and video generation.
The model-provider layer is also important. The Register reported that Vercel’s AI SDK and AI Gateway support multiple model providers. Vercel’s enterprise post positions AI Gateway as a way to connect applications to models while retaining options such as provider fallbacks.
That approach does not eliminate reliance on external AI model vendors, but it can reduce lock-in at the application layer. If an application is built against Vercel’s SDK and gateway abstractions, teams may have more flexibility to route requests across providers or change model choices without rewriting every part of the application.
Vercel’s enterprise announcement also highlights Passport, Connect, managed users, and bring-your-own-cloud deployment on AWS. The company says these pieces are meant to keep agents behind company access and security boundaries.
Passport and Connect are presented as governance and integration mechanisms for enterprise AI use. The Register characterized Passport as part of Vercel’s attempt to address “shadow AI,” where employees or teams adopt AI tools outside approved corporate controls.
Bring-your-own-cloud on AWS is another key part of the pitch. Vercel says BYOC allows customers to run workloads in their own AWS environment while using Vercel’s platform. For regulated or security-conscious organizations, this can be relevant when internal policies require tighter control over where workloads run and how infrastructure is managed.
Taken together, Vercel’s announcements suggest a strategy focused less on owning a frontier model and more on owning the developer and deployment layer around AI applications. The company is combining its existing strengths in web application development with newer agent infrastructure, model routing, identity, and governance tools.
That is a practical enterprise bet. Many companies are already experimenting with AI assistants and agents, but production adoption depends on familiar software concerns: authentication, authorization, observability, approvals, deployment, reliability, and vendor management. Vercel’s materials argue that these concerns should be handled as part of the application platform, not bolted on after a prototype succeeds.
The announcements do not prove that Vercel’s stack will become the default way enterprises build agents. Large organizations already use a mix of cloud platforms, internal tooling, model APIs, and orchestration frameworks. But the company is making a clear case that AI agents should be governed like enterprise software, with controls for users, tools, models, and infrastructure from the start.
For teams already using Next.js or Vercel, the new offerings could provide an incremental path into AI application development. For enterprises evaluating agent platforms more broadly, Vercel’s launch adds another option in a fast-moving market where the durable value may lie not only in model access, but in the systems that make AI applications safe enough to run inside a company.
Vercel announced an enterprise platform for AI apps and agents that combines its new eve framework with governance, identity, deployment, and model routing tools.
Eve is Vercel’s new agent framework Vercel introduced eve as an open source agent framework and described it as a production oriented foundation for building AI agents.
According to Vercel, eve includes durable execution, sandboxed compute, human approvals, evaluations, AI Gateway fallbacks, and a file based structure for defining agents.
Continue reading
