Vercel Sandbox Adds Persistent Drives for AI Agents to Retain Memory

VercelVercel

Vercel Sandbox now supports persistent drives in private beta. This separates storage from compute, allowing AI agents to retain their workspaces and "memories" across disposable sandbox instances. This is crucial for building reliable, multi-step agentic systems that need to maintain state.

Vercel Sandbox now supports drives in private beta, providing persistent, attachable storage with a lifecycle independent from any sandbox. This separates storage from compute, allowing users to create a drive once and mount it at a configurable path when starting an isolated execution environment (sandbox). When the sandbox stops, the drive remains available to attach to a later sandbox.
Availability
Private beta
Access
Pro and Enterprise teams
Mounting
One read-write mount per sandbox (during beta)
Use Case
Keeping agent workspaces across disposable sandboxes
Current Limitation
Not for production data (during beta)

This capability is crucial for building reliable AI agent systems, as agents can now retain their workspaces, cloned repositories, dependencies, and build outputs across disposable environments. The ability to maintain state across sessions addresses a key challenge in agentic coding, where agents need to remember past actions and data. Other platforms are also focusing on persistent environments for Cloudflare Sandboxes and dedicated agent memory solutions.

The drives are in private beta for Pro and Enterprise teams, capped at one read-write mount per sandbox and not meant for production data yet. The payoff is an agent workspace that survives across disposable sandboxes — the same problem Anthropic tackled in its decoupled agent architecture.

View the full update on vercel.com
Vercel Developers
Vercel Developers
@vercel_dev
X

Sandbox now supports drives in early access ▪︎ Storage separated from compute ▪︎ Mount on any path with 𝙳𝚛𝚒𝚟𝚎.𝚐𝚎𝚝𝙾𝚛𝙲𝚛𝚎𝚊𝚝𝚎 ▪︎ Keep agent memories across sandboxes Sign up ↓ https://t.co/o0ODXM6MZa

1retweets9likes
View on X

Still wondering? A few quick answers below.

Drives for Vercel Sandbox are persistent, attachable storage units that operate independently from the lifecycle of any specific sandbox. This means you can create a drive once and then mount it to different sandboxes as needed, ensuring your data persists even when a sandbox stops.

Drives allow AI agents to retain their "memories" or workspaces across multiple sandbox sessions. This is essential for agentic coding and complex tasks where agents need to maintain state, access cloned repositories, manage dependencies, and keep build outputs without losing information when their execution environment resets.

The drives for Vercel Sandbox feature is currently available in private beta. Access is limited to Pro and Enterprise teams who sign up for the waitlist. Users need to install the beta SDK or CLI to use this functionality.

During the private beta, a drive can only be mounted in read-write mode by one sandbox at a time. Additionally, Vercel advises against using Sandbox Drives for production data while the feature is in its private beta phase.

Every HeadsUpAI update is written based on its original source and reviewed before it's published. Read our editorial standards →

Share this update