Vercel Connect Launches to Secure Agent Access with Short-Lived Tokens.

Vercel has launched Vercel Connect, a new capability designed to simplify and secure how apps and agents access external data and systems. The platform uses short-lived tokens with precise scopes to reduce risk when integrating with third-party services, a growing concern as AI agents become more autonomous.

The announcement comes as Vercel continues to build out its Agent Stack, following the recent unveiling of Eve, an open-source framework for production AI agents. Vercel Connect addresses a critical gap: how to give agents the access they need to external APIs, databases, and SaaS tools without over-permissioning or leaving credentials exposed.

Traditional API keys and long-lived credentials pose security challenges, especially when agents operate autonomously across multiple systems. A leaked key can expose entire data stores or allow unauthorized actions. Vercel Connect mitigates this by issuing tokens that expire quickly and carry only the permissions required for a specific task.

Each token is scoped to a narrow set of actions, such as reading from a specific database table or posting to a single endpoint. This principle of least privilege means that even if a token is compromised, the damage is contained. The tokens are issued dynamically as agents need them, reducing the attack surface compared to static credentials stored in environment variables.

Vercel Connect is part of the broader Agent Stack, which also includes the AI SDK, AI Gateway, Sandbox, and Workflows. Together, these tools provide the infrastructure for building, running, and securing AI agents at scale. Vercel Connect sits between agents and external services, handling authentication and authorization without requiring developers to implement custom token management logic.

Developers can configure which services their agents can access and define the specific scopes allowed. When an agent needs to interact with an external system, Vercel Connect generates a token on the fly, the agent uses it to complete the task, and the token expires shortly after. This workflow is transparent to the agent, which receives credentials through a standard interface without needing to know about the underlying token lifecycle.

Vercel Connect is currently in early access, with a focus on enterprise customers who need to integrate agents with internal systems and third-party SaaS platforms. The company has emphasized security and compliance as core design principles, aligning with the capabilities announced in Vercel's enterprise agent platform, which includes built-in security controls and support for running agents in customer-owned AWS accounts.

Companies building AI-powered workflows that span multiple services will likely find Vercel Connect particularly useful. Examples include agents that read data from a CRM, process it with a model, and write results to a spreadsheet, or bots that monitor support tickets and create tasks in project management tools. In each case, Vercel Connect ensures the agent has the access it needs without broad, persistent credentials that could be misused.

The launch reflects a broader industry shift toward treating agent security as a first-class concern, rather than an afterthought. As agents take on more responsibilities, from customer service to internal automation, the stakes for credential management rise. Vercel Connect offers a path forward that balances functionality with containment, giving builders the tools to deploy agents confidently.