Agent workspaces and file-based memory can persist across sandbox runs, and forking a Drive lets parallel workers start from identical files with isolated changes.
Key takeaways · AI-distilled
Drives are generally available for Vercel sandboxAn isolated environment where AI-generated code or agent actions run without being able to touch anything real.Full definition → on the Hobby, Pro, and Enterprise plans.
A fork inherits the source Drive's data, region, and maximum size, then records only its own changes, leaving the original and sibling forks untouched.
Beyond AI agentAn AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.Full definition → memory, Vercel suggests Drives for sharing datasets, models, dependencies, and build artifacts across sandboxes.
In the SDK example, each worker calls drive.fork with its own name, mounts the fork at a path such as /data when creating a Sandbox, and deletes the fork after the session.
Terms in this piece · Glossary
sandbox — An isolated environment where AI-generated code or agent actions run without being able to touch anything real.
AI agent — An AI system that doesn't just answer once but works toward a goal in a loop — taking actions, reading the results, and deciding what to do next.
