@vercel/sandbox

Vercel Sandbox

Vercel Sandbox allows you to run arbitrary code in isolated, ephemeral Linux VMs. View the documentation here.

Packages

• @vercel/sandbox (this package) - The SDK for programmatic access to Vercel Sandbox. Source | Documentation
• sandbox - The CLI for interacting with Vercel Sandbox from the command line. Source | Documentation

What is a sandbox?

A sandbox is an isolated Linux system for your experimentation and use. Internally, it is a Firecracker MicroVM that is powered by the same infrastructure that powers 2M+ builds a day at Vercel.

Getting started

To get started using Node.js 22+, create a new project:

mkdir my-sandbox-app && cd my-sandbox-app
npm init -y
vercel link

Pull your authentication token:

vercel env pull

Install the Sandbox SDK:

pnpm i @vercel/sandbox

Create a index.mts file:

import { Sandbox } from "@vercel/sandbox";
import { setTimeout } from "timers/promises";
import { spawn } from "child_process";

async function main() {
  const sandbox = await Sandbox.create({
    source: {
      url: "https://github.com/vercel/sandbox-example-next.git",
      type: "git",
    },
    resources: { vcpus: 4 },
    ports: [3000],
    runtime: "node24",
    name: "vercel-sandbox-example",
  });
  console.log(`Sandbox ${sandbox.name} created`);

  console.log(`Installing dependencies...`);
  const install = await sandbox.runCommand({
    cmd: "npm",
    args: ["install", "--loglevel", "info"],
    stderr: process.stderr,
    stdout: process.stdout,
  });

  if (install.exitCode !== 0) {
    console.log("installing packages failed");
    process.exit(1);
  }

  console.log(`Starting the development server...`);
  await sandbox.runCommand({
    cmd: "npm",
    args: ["run", "dev"],
    stderr: process.stderr,
    stdout: process.stdout,
    detached: true,
  });

  await setTimeout(500);
  spawn("open", [sandbox.domain(3000)]);
}

main().catch(console.error);

Run it:

node --experimental-strip-types --env-file .env.local index.mts

This will:

• Start a sandbox, seeding it with a git repository.
• Install dependencies.
• Run a next dev server
• Open it in your browser

All while streaming logs to your local terminal.

Sandboxes are persistent by default. To resume a sandbox with its previous state:

Create a resume.mts file:

import { Sandbox } from "@vercel/sandbox";
import { setTimeout } from "timers/promises";
import { spawn } from "child_process";

async function main() {
  const sandbox = await Sandbox.get({
    name: "vercel-sandbox-example",
  });
  console.log(`Sandbox ${sandbox.name} resumed`);

  console.log(`Starting the development server...`);
  await sandbox.runCommand({
    cmd: "npm",
    args: ["run", "dev"],
    stderr: process.stderr,
    stdout: process.stdout,
    detached: true,
  });

  await setTimeout(500);
  spawn("open", [sandbox.domain(3000)]);
}

main().catch(console.error);

Run it:

node --experimental-strip-types --env-file .env.local resume.mts

Authentication

Vercel OIDC token

The SDK uses Vercel OIDC tokens to authenticate whenever available. This is the most straightforward and recommended way to authenticate.

When developing locally, you can download a development token to .env.local using vercel env pull. After 12 hours the development token expires, meaning you will have to call vercel env pull again.

In production, Vercel manages token expiration for you.

Access token

If you want to use the SDK from an environment where VERCEL_OIDC_TOKEN is unavailable, you can also authenticate using an access token:

• Go to your team settings, and copy the team ID.
• Go to a project's settings, and copy the project ID.
• Go to your Vercel account settings and create a token. Make sure it is scoped to the team ID from the previous step.

Set your team ID, project ID, and token to the environment variables VERCEL_TEAM_ID, VERCEL_PROJECT_ID, and VERCEL_TOKEN. Then pass these to the create method:

const sandbox = await Sandbox.create({
  teamId: process.env.VERCEL_TEAM_ID!,
  projectId: process.env.VERCEL_PROJECT_ID!,
  token: process.env.VERCEL_TOKEN!,
  source: {
    url: "https://github.com/vercel/sandbox-example-next.git",
    type: "git",
  },
  resources: { vcpus: 4 },
  // Defaults to 5 minutes. The maximum is 24 hours for Pro/Enterprise, and 45 minutes for Hobby.
  timeout: ms("5m"),
  ports: [3000],
  runtime: "node24",
});

Workflow DevKit integration

Sandbox and CommandFinished support serialization with the Workflow DevKit. When a sandbox instance crosses a step boundary the SDK serializes sandbox metadata and routes, then rehydrates synchronously from that snapshot. Deserialized instances lazily recreate an API client using OIDC or environment credentials when needed.

Limitations

• Max resources: 8 vCPUs. You will get 2048 MB of memory per vCPU.
• Sandboxes have a maximum runtime duration of 24 hours for Pro/Enterprise and 45 minutes for Hobby, with a default of 5 minutes. This can be configured using the timeout option of Sandbox.create().

System

The base system is an Amazon Linux 2023 system with the following additional packages installed.

bind-utils
bzip2
findutils
git
gzip
iputils
libicu
libjpeg
libpng
ncurses-libs
openssl
openssl-libs
procps
tar
unzip
which
whois
zstd

• The node24 and node22 images ship Node runtimes under /vercel/runtimes/node{22,24}.
• The python3.13 image ships a Python 3.13 runtime under /vercel/runtimes/python.
• User code is executed as the vercel-sandbox user.
• /vercel/sandbox is writable.

Custom images

Instead of a stock runtime, you can start a sandbox from a Vercel Container Registry (VCR) image with the image option:

import { Sandbox } from "@vercel/sandbox";

const sandbox = await Sandbox.create({
  image: "my-repo:v1",
});

The image option accepts a repository in the sandbox's project, with an optional tag or digest. A bare repository name resolves to the latest tag. You can also pass a fully-qualified VCR URL:

await Sandbox.create({ image: "my-repo" }); // latest tag
await Sandbox.create({ image: "my-repo:v1" }); // specific tag
await Sandbox.create({ image: "my-repo@sha256:..." }); // specific digest
await Sandbox.create({
  image: "vcr.vercel.com/my-team/my-project/my-repo:v1", // fully-qualified
});

Sudo access

The nodeX and python3.13 images allow users to run commands as root. This can be used to install packages and system tools:

import { Sandbox } from "@vercel/sandbox";

const sandbox = await Sandbox.create();
await sandbox.runCommand({
  cmd: "dnf",
  args: ["install", "-y", "golang"],
  sudo: true,
});

Sandbox runs sudo in the following configuration:

• HOME is set to /root – Executed commands will source root's configuration files (e.g. .gitconfig, .bashrc, etc).
• Environment variables are not reset before executing the command.
• PATH is left unchanged – sudo won't change the value of PATH, so local or project-specific binaries will still be found.

Both these images are based on Amazon Linux 2023. The full package list is available here.

Multi-user

Sandboxes support creating isolated Linux users with their own home directories, file permissions, and process ownership. This is useful for multi-agent workflows where each agent needs its own workspace, or for simulating multi-user environments.

Note: The sandbox image must have /bin/bash installed. It is the login shell for created users and is used to wrap commands that run as a user. The stock Vercel Sandbox images include it.

Creating users

import { Sandbox } from "@vercel/sandbox";

const sandbox = await Sandbox.create();

// Creates /home/alice with isolated permissions
const alice = await sandbox.createUser("alice");

alice.username; // "alice"
alice.homeDir;  // "/home/alice"

createUser sets up:

• A Linux user with /bin/bash as the default shell
• A home directory at /home/<username> group-owned by the sandbox's default user group with 770 permissions

Running commands as a user

All commands run as the user by default, with the working directory set to their home:

const alice = await sandbox.createUser("alice");

const whoami = await alice.runCommand("whoami");
await whoami.stdout(); // "alice\n"

const pwd = await alice.runCommand("pwd");
await pwd.stdout(); // "/home/alice\n"

You can pass environment variables, override the working directory, or use the full RunCommandParams interface:

// Environment variables
await alice.runCommand({
  cmd: "node",
  args: ["-e", "console.log(process.env.API_KEY)"],
  env: { API_KEY: "secret" },
});

// Custom working directory
await alice.runCommand({ cmd: "ls", cwd: "/tmp" });

// Detached mode for long-running processes
const server = await alice.runCommand({
  cmd: "node",
  args: ["server.js"],
  detached: true,
});

To escalate to root, pass sudo: true:

await alice.runCommand({
  cmd: "dnf",
  args: ["install", "-y", "git"],
  sudo: true,
});

File operations

writeFiles, readFile, readFileToBuffer, and mkDir all resolve relative paths against the user's home directory. Written files are owned by the user:

const alice = await sandbox.createUser("alice");

// Writes to /home/alice/app.js, owned by alice:alice
await alice.writeFiles([
  { path: "app.js", content: Buffer.from('console.log("hi")') },
]);

// Read it back
const buf = await alice.readFileToBuffer({ path: "app.js" });
buf?.toString(); // 'console.log("hi")'

// Stream reads
const stream = await alice.readFile({ path: "app.js" });

// Create directories owned by the user
await alice.mkDir("projects/my-app");

// Absolute paths also work
await alice.writeFiles([
  { path: "/tmp/output.txt", content: Buffer.from("data") },
]);

File isolation

Users cannot access each other's home directories:

const alice = await sandbox.createUser("alice");
const bob = await sandbox.createUser("bob");

await alice.writeFiles([
  { path: "secret.txt", content: Buffer.from("alice only") },
]);

// Bob cannot read, list, or write to alice's home
const cat = await bob.runCommand({
  cmd: "cat",
  args: ["/home/alice/secret.txt"],
});
cat.exitCode; // non-zero — Permission denied

The SDK can read all users' files because home directories are group-owned by the sandbox's default user group. Both SandboxUser methods and direct sandbox methods work:

// Via SandboxUser (relative paths resolve to home dir)
const buf = await alice.readFileToBuffer({ path: "secret.txt" });
buf?.toString(); // "alice only"

// Via sandbox directly (absolute path required)
const buf2 = await sandbox.readFileToBuffer({ path: "/home/alice/secret.txt" });
buf2?.toString(); // "alice only"

Groups and shared directories

Create groups to let users collaborate through a shared directory:

const devs = await sandbox.createGroup("devs");
devs.sharedDir; // "/shared/devs"

await sandbox.addUserToGroup("alice", "devs");
await sandbox.addUserToGroup("bob", "devs");

// Alice writes to the shared directory
await alice.runCommand({
  cmd: "bash",
  args: ["-c", 'echo "spec v2" > /shared/devs/spec.txt'],
});

// Bob can read it — files inherit group ownership via setgid
const spec = await bob.runCommand({
  cmd: "cat",
  args: ["/shared/devs/spec.txt"],
});
await spec.stdout(); // "spec v2\n"

// Non-members are blocked
const charlie = await sandbox.createUser("charlie");
const ls = await charlie.runCommand({ cmd: "ls", args: ["/shared/devs"] });
ls.exitCode; // non-zero — Permission denied

Shared directories use setgid (2770), so files created inside them automatically inherit the group. All group members get read/write access.

Convenience methods are available on SandboxUser:

await alice.addToGroup("devs");
await alice.removeFromGroup("devs");

Using asUser for existing users

If a user already exists (e.g., from a snapshot or manual creation), use asUser to get a handle without re-creating:

const existing = sandbox.asUser("bob");
await existing.runCommand("whoami"); // "bob"

Username validation

Usernames and group names must match /^[a-z_][a-z0-9_-]*$/ and be at most 32 characters. Invalid names throw an error immediately:

sandbox.asUser("Alice");        // throws — uppercase
sandbox.asUser("user name");    // throws — space
sandbox.asUser("$(whoami)");    // throws — special characters
sandbox.asUser("a".repeat(33)); // throws — too long

Multi-agent example

const sandbox = await Sandbox.create();

// Each agent gets its own isolated workspace
const researcher = await sandbox.createUser("researcher");
const coder = await sandbox.createUser("coder");
const reviewer = await sandbox.createUser("reviewer");

// Shared workspace for collaboration
await sandbox.createGroup("project");
await sandbox.addUserToGroup("researcher", "project");
await sandbox.addUserToGroup("coder", "project");
await sandbox.addUserToGroup("reviewer", "project");

// Researcher writes findings to shared dir
await researcher.runCommand({
  cmd: "bash",
  args: ["-c", 'echo "API spec v2" > /shared/project/spec.txt'],
});

// Coder reads spec, writes code in their own home
const spec = await coder.runCommand({
  cmd: "cat",
  args: ["/shared/project/spec.txt"],
});
await coder.writeFiles([
  { path: "app.js", content: Buffer.from(`// ${await spec.stdout()}`) },
]);

// Reviewer can read the shared spec but not coder's private files
const blocked = await reviewer.runCommand({
  cmd: "cat",
  args: ["/home/coder/app.js"],
});
blocked.exitCode; // non-zero — isolation enforced

Authors

This library is created by Vercel team members, with contributions from the Open Source Community welcome and highly appreciated.