Vercel Drives
A persistent storage volume offering from Vercel that separates the file layer from the execution layer, making an AI agent's files independent of the execution environment. It supports mounting across Sandboxes, shared reads and forking.
Timeline (last 90 days)
Vercel Drives is now generally available for Vercel Sandbox, covering all three tiers: Hobby, Pro, and Enterprise.
The two main uses of Vercel Drives are: saving an Agent's workspace or file-based memory to preserve state across runs, and sharing datasets, model weights, dependencies, and build artifacts across sandboxes.
Copy-on-Write is the key design idea behind the Fork mechanism: a forked Drive initially retains only lightweight references pointing to the source data, and only when a Fork actually modifies a file are the changed data blocks copied.
Fork is suited to scenarios of "starting from the same initial workspace and producing different results," not to high-write workloads that require continuously writing large amounts of new data after forking.
Vercel Drives takes the "files as memory" route, where the Agent writes the workspace, notes, and intermediate artifacts as ordinary files, and the next run can simply mount the same Drive to read them, with no need to introduce an additional memory service.
File-based memory is inferior to dedicated databases in retrieval efficiency and concurrent-write consistency, and still needs to be evaluated in scaled scenarios.
Vercel Drives' actual performance, concurrency limits, and cross-region behavior still need to be verified under real workloads, and teams are advised to try it first in small-scale parallel Worker scenarios before deciding whether to use it as a core storage layer.
Vercel Drives can be created and mounted in two ways: via the CLI (sandbox drives get-or-create, sandbox create --mount) and via the SDK (the Sandbox and Drive classes provided by @vercel/sandbox).
Vercel Drives is a persistent storage solution provided by Vercel that makes Agent files independent of the execution environment, and is currently generally available (GA) on Vercel Sandbox.
Vercel Drives supports shared read scenarios: multiple Agents can mount read-only copies, with each keeping its results in its own space without interfering with others.
6 more timeline events
All Facts (16)
The Fork in Vercel Drives uses a copy-on-write mechanism, with the fork operation costing close to O(1) and completing instantly regardless of how large the original Drive is.
60%UnverifiedA Drive in Vercel Drives is a persistent volume independent of the Sandbox; the Sandbox associates it with a file system path via mounting, with reads and writes below the mount point landing on the Drive and those outside being temporary storage.
60%UnverifiedThe two main uses of Vercel Drives are: saving an Agent's workspace or file-based memory to preserve state across runs, and sharing datasets, model weights, dependencies, and build artifacts across sandboxes.
50%UnverifiedCopy-on-Write is the key design idea behind the Fork mechanism: a forked Drive initially retains only lightweight references pointing to the source data, and only when a Fork actually modifies a file are the changed data blocks copied.
50%UnverifiedVercel Drives' actual performance, concurrency limits, and cross-region behavior still need to be verified under real workloads, and teams are advised to try it first in small-scale parallel Worker scenarios before deciding whether to use it as a core storage layer.
50%UnverifiedVercel Drives takes the "files as memory" route, where the Agent writes the workspace, notes, and intermediate artifacts as ordinary files, and the next run can simply mount the same Drive to read them, with no need to introduce an additional memory service.
50%UnverifiedFork is suited to scenarios of "starting from the same initial workspace and producing different results," not to high-write workloads that require continuously writing large amounts of new data after forking.
50%UnverifiedFile-based memory is inferior to dedicated databases in retrieval efficiency and concurrent-write consistency, and still needs to be evaluated in scaled scenarios.
50%UnverifiedVercel Drives supports shared read scenarios: multiple Agents can mount read-only copies, with each keeping its results in its own space without interfering with others.
50%UnverifiedMultiple Sandboxes can mount the same Drive simultaneously, and switching at runtime requires no additional data migration.
50%UnverifiedVercel Drives supports a Fork mechanism, where an Agent can fork a Drive, each fork starting from the original content and having an independent writable workspace, and forks are not merged back into the original Drive. The Fork feature is generally available.
50%UnverifiedVercel Drives has a default capacity of 1 TiB, 1 GiB for the Hobby plan, and can be expanded up to a maximum of 16 TiB.
50%UnverifiedVercel Drives is open to all plans of Vercel Sandbox.
50%UnverifiedVercel Drives can be created and mounted in two ways: via the CLI (sandbox drives get-or-create, sandbox create --mount) and via the SDK (the Sandbox and Drive classes provided by @vercel/sandbox).
50%UnverifiedVercel Drives is now generally available for Vercel Sandbox, covering all three tiers: Hobby, Pro, and Enterprise.
50%UnverifiedVercel Drives is a persistent storage solution provided by Vercel that makes Agent files independent of the execution environment, and is currently generally available (GA) on Vercel Sandbox.
50%