Vercel Unveils Agent Stack at Ship 26 London with Eve Framework and Security Tools.

Vercel announced a suite of agent-focused products at Ship 26 London on June 17, positioning the company deeper into the AI development stack. The releases include eve, an open-source framework for building production agents, alongside enterprise security and connectivity tools designed to help teams deploy and manage AI systems at scale.

The announcement marks a significant expansion of Vercel's platform beyond its core frontend deployment offering. Chief Product Officer Tom Occhino introduced the Agent Stack as a comprehensive set of building blocks for creating and shipping production-grade agents.

Eve is Vercel's answer to the fragmented landscape of agent development. The framework provides developers with standardized patterns for building, running, and scaling AI agents, similar to how Next.js brought structure to React applications.

Key features include:

• Built-in orchestration for multi-step agent workflows
• Native support for tool calling and external integrations
• Type-safe agent definitions
• Local development environment with hot reloading

The framework integrates with Vercel's existing AI SDK and can deploy to Vercel's infrastructure or run on-premises. Developers can define agent behaviors using TypeScript and leverage Vercel's Sandbox environment for secure code execution.

Vercel Connect addresses a critical security challenge in agent deployment: giving AI systems access to enterprise tools without exposing long-lived credentials. The service supports integrations with Slack, GitHub, Snowflake, Salesforce, Notion, Linear, and any Model Context Protocol (MCP) URL.

Connect issues short-lived tokens scoped to specific actions, following security principles similar to Vercel Connect announced earlier this year. This approach limits blast radius if an agent is compromised and provides granular audit logs for compliance teams.

The integration process requires minimal configuration. Developers authenticate their services once through Vercel's dashboard, then reference connections in agent code. Connect handles token refresh, rate limiting, and error recovery automatically.

Vercel Agent is the company's first internal product built on the Agent Stack. It reviews pull requests, investigates deployment anomalies, and proposes fixes directly in GitHub.

The system analyzes code changes against deployment metrics, identifying performance regressions, security vulnerabilities, and accessibility issues. When problems are detected, Vercel Agent can open follow-up PRs with suggested fixes.

This capability builds on Vercel's enterprise agent platform, extending automated workflows beyond deployment into code quality management. The tool integrates with existing CI/CD pipelines and respects branch protection rules.

Vercel Passport solves access control for preview and production deployments by integrating with enterprise identity providers like Okta and Auth0. Teams can require authentication before users access specific environments, useful for staging sites, internal tools, or customer-specific deployments.

Passport supports SAML and OIDC protocols. Configuration happens at the project level in Vercel's dashboard. Once enabled, unauthenticated visitors see a login screen that redirects to the configured identity provider.

The feature complements Vercel's existing authentication options and provides audit trails for compliance requirements. Access policies can be scoped by team, role, or custom attributes from the IdP.

Eve is open source and available on GitHub. Vercel Connect, Vercel Agent, and Vercel Passport are rolling out to Enterprise customers, with broader availability planned for later this year.

Developers can access eve documentation and starter templates through Vercel's CLI. The framework supports deployment to Vercel's edge network and standard Node.js environments.

Vercel plans additional announcements at upcoming Ship events in Berlin on June 25 and New York City on June 30. The company has been steadily expanding its AI capabilities, recently removing SDK key requirements for feature flags to simplify platform-native integrations.