Vercel Sandbox now supports custom tags in beta, letting teams label and filter isolated environments at scale. Each sandbox accepts up to five tags. The feature targets three concrete problems: attributing sandbox usage to specific teams, separating staging from production environments, and isolating sandboxes per customer in multi-tenant platforms.
Tags are mutable. You can reassign ownership, promote a sandbox from staging to production, or mark it for cleanup without recreating the environment. That mutability is the detail worth reading about: the filtering API lets dashboards, cleanup scripts, and routing logic query all sandboxes matching a given tag combination, which matters when you are managing hundreds of ephemeral agent or code-gen environments.
The beta requires upgrading the @vercel/sandbox SDK and CLI packages. Full implementation details, including the tag filtering API shape and multi-tenant attribution patterns, are in the official documentation. If you are running AI agents at scale or billing customers against isolated workspaces, the use case breakdown alone is worth the click.
[READ ORIGINAL →]