Scale AI's AgentEnv adds Vercel Sandbox provider for running agent evals
Vercel announced that Scale AI's AgentEnv, a framework for testing agents against tasks and scoring their results, now supports running evaluations on Vercel Sandbox. The Vercel provider ships in AgentEnv 0.9.1334 and later and runs the agent and its task environment in separate Sandboxes.
Why it matters
Builders evaluating AI agents can run AgentEnv evaluations on Vercel Sandbox without managing their own infrastructure, with the agent and task environment isolated in separate Sandboxes. The announcement describes setup through an example repository and the Vercel CLI login.
Who should care
Developers building or testing AI agents, teams running agent evaluations, and engineers using Vercel Sandbox.
What you can do
Clone the AgentEnv examples repository, follow the setup guide, and run an evaluation on Vercel Sandbox using your existing Vercel CLI login to inspect task scores and recorded agent actions.