Vercel's new open-source framework eve treats AI agents as simple file directories

Vercel has launched eve, an open-source framework for building, running, and scaling AI agents, now available in public preview. The framework aims to simplify agent development by treating an agent as a simple directory of files, with the smallest agent requiring just two files: a model and instructions. Adding tools, skills, channels, or schedules is done by adding files, which eve automatically wires in at build time without boilerplate, according to Vercel. "Adding tools, skills, channels, or schedules is done by adding files, which eve automatically wires in at build time without boilerplate." This file-based approach is designed to eliminate the need for developers to hand-roll common infrastructure. Vercel reported that users can scaffold and start a new agent with a single command that installs dependencies, scaffolds the project, and starts a dev server, making the initial setup process straightforward. The framework includes built-in features like durable execution, sandboxed compute, approvals, channels, tracing, and evals. Vercel draws a parallel to how Next.js standardized web development, positioning eve as a similar standardizing force for agent development. By providing these capabilities out of the box, the company aims to reduce the complexity of building production-ready AI agents. "It draws a parallel to how Next.js standardized web development, positioning eve as a similar standardizing force for agent development." With eve, Vercel is targeting developers who want to focus on agent logic rather than infrastructure setup. The framework's open-source nature and public preview status invite community experimentation and feedback, potentially accelerating adoption among teams building AI-powered applications.

2 sources•20 JunRead the Baker's Take→