Vercel deepens agentic AI infrastructure with eve framework, Agent Stack, and enterprise identity controls

Vercel is moving beyond front-end hosting to become the infrastructure layer for agentic AI, introducing a coordinated set of tools at its Ship conference that address the full lifecycle of building, deploying, and managing AI agents in production. For teams already using Next.js or Vercel's AI SDK, these launches provide a unified path from agent development to enterprise-grade governance.

What happened

Vercel announced five major products at Ship 2026 to deepen its agentic AI infrastructure platform. The company reported that fewer than 3% of deployments on its platform were triggered by coding agents at the start of the year; now agents account for more than half of all commits. Token volume through the Vercel AI Gateway grew from roughly 2 million to 20 million over the same period.

The new offerings include:

Vercel Services provides project-package deployment with a single commit generating a single URL for an entire application, supporting Next.js front ends and Python back ends on Vercel. It enters beta on July 1.

The Agent Stack bundles existing Vercel AI tooling: AI Gateway, Sandbox, Workflow SDK, and Chat SDK. It is complemented by Vercel Connect, which replaces long-lived credentials with short-lived tokens tied to specific people or agents, along with full audit trails. Connect launches with integrations for Slack, GitHub, Snowflake, Salesforce, Notion, and Linear, with connectors available for OAuth or APIs.

eve is an open-source, opinionated agent framework that Vercel describes as a scaffold for building, running, and scaling agents. The company says it uses eve internally and is now releasing it for external experimentation.

Vercel Agent proactively helps developers manage production applications and agent infrastructure by viewing traffic, traces, and alerts, then autonomously investigating issues and recommending fixes for human approval. It is available in beta now.

Vercel for Enterprise Apps and Agents packages the full toolset with identity and access controls, including Enterprise Managed Users for global management of Vercel and v0 users, and Vercel Passport for integrating with a company's identity provider. Organizations can optionally deploy within their own AWS account.

Why AI builders should care

The shift from human-triggered deployments to agent-triggered ones is already happening on Vercel's platform. For teams building AI agents, the key takeaway is that Vercel is making it possible to ship a full-stack application and its supporting agent infrastructure from a single commit, with built-in short-lived token management and audit trails.

For builders using Next.js, the ability to deploy Python back ends alongside front ends in one project removes a common friction point. The eve framework gives teams a starting point for structuring agents, and because it is open source, it can be customized or audited.

Enterprise teams that need to deploy agents inside their own AWS account and integrate with existing identity providers now have a formal mechanism through Vercel for Enterprise Apps and Agents. The inclusion of Vercel Passport means internal apps and agents can live behind a company's IdP by default.

Practical implications

For teams already on Vercel, the Agent Stack and Connect reduce the need to cobble together separate tools for token management, sandboxing, and audit logging. The Workflow SDK and Chat SDK can be used together to build agent workflows that trigger automated actions or handle user-facing chat interfaces.

The Vercel Agent (in beta) is worth testing for teams running production workloads: it can autonomously investigate incidents and suggest fixes, which could reduce mean time to resolution for common operational issues.

For teams evaluating agent frameworks, eve offers an opinionated alternative to LangChain or CrewAI, tightly integrated with Vercel's deployment and monitoring infrastructure. However, early adopters should treat it as experimental given the open-source release is new.

Caveats

Most of these products are in beta: Vercel Services enters beta on July 1, Vercel Agent is in beta now, and eve is released as open source but likely in early stages. The article does not provide pricing for any of these new offerings, so cost implications for enterprise use are unclear.

The enterprise controls (Vercel for Enterprise Apps and Agents) require an AWS account for self-hosted deployment; teams using other cloud providers may need to adapt. Vercel Connect integrations are limited to a small set of popular services at launch, though OAuth and API connectors provide extensibility.

As with any vendor announcement, particularly one centered on agentic AI, claims about deployment statistics and token growth should be treated as directional. The 50% agent-triggered deployment figure reflects Vercel's platform mix and may not generalize.