Vercel Sandbox Now Runs on Vercel Managed Images.

Vercel is replacing its Sandbox runtimes with a new set of managed, open-source base images, giving developers a more standardized environment for running generated code, previews and tests. Starting with version 3 of the Sandbox SDK, newly created sandboxes will default to a universal image maintained by Vercel rather than relying on the previous runtime system.

The company announced the change in its changelog, positioning Vercel Managed Images, or VMI, as both a ready-made default and a foundation developers can extend. The source for each image is available in the public vercel/sandbox repository, which should make it easier for teams to inspect what is installed and understand changes between versions.

The main new default is vercel/sandbox/universal:latest. It includes Node.js, Python, common coding agents and standard system utilities, covering many of the dependencies needed for AI-assisted development and code execution. Vercel’s aim is to reduce the number of teams that need to create a custom image or install packages every time a sandbox starts.

The update introduces several practical changes:

• Managed base images: Vercel now maintains the underlying images used by Sandbox.
• A broader universal default: Node.js, Python, coding agents and frequently used utilities are included.
• Open-source definitions: Developers can review and extend the images through Vercel’s public repository.
• A new operating system: The universal image moves the default environment from Amazon Linux to Ubuntu.
• Deprecated runtimes: The previous Sandbox runtime model is being retired in favor of managed images.

Ubuntu is a meaningful change even if it is largely invisible to users. It is common across cloud development environments and has a large package ecosystem, which can make dependency instructions and troubleshooting more portable. However, code or setup scripts that assumed Amazon Linux behavior may require review, particularly where they depend on system packages, shell behavior or operating-system-specific paths.

Sandbox infrastructure is increasingly important for products that let AI agents write and execute code. A generated application may need to install dependencies, run a development server, execute tests and expose a preview, all without gaining access to the host system. The more predictable that environment is, the fewer failures developers should encounter because one sandbox differs from another.

For small teams, the immediate benefit is reduced image maintenance. A preconfigured environment can remove the need to assemble and update a base image solely to support common JavaScript and Python workflows. Bundling standard tools also avoids repeated installation work at startup, which is intended to improve build performance while making sandbox behavior more consistent.

There is still room for customization. Because the managed images can be used as-is or extended, teams with specialized dependencies are not locked into the universal configuration. That balance matters for businesses using Sandbox as infrastructure inside code-generation products, internal tools or customer-facing development environments. They can begin with Vercel’s maintained baseline, then add only the packages their workload requires.

Developers adopting Sandbox SDK version 3 should nevertheless treat the change as a runtime migration, not just a renamed default. Existing scripts should be checked against Ubuntu, while custom images and package installation steps may be candidates for simplification. Teams that require reproducible environments should also evaluate whether to pin a versioned managed image rather than depend indefinitely on the mutable latest tag.

Vercel has not disclosed a separate price for Managed Images or a final removal date for deprecated Sandbox runtimes in the supplied announcement. The clearest immediate milestone is SDK version 3, where new sandboxes begin using the universal managed image automatically.

What is Vercel Sandbox?

Vercel Sandbox is an isolated environment for safely running code, tests and application previews. With version 3 of its SDK, newly created sandboxes default to a Vercel Managed Image containing common languages, coding agents and utilities.

What is included in the new universal image?

The vercel/sandbox/universal:latest image includes Node.js, Python, common coding agents and standard utilities. It uses Ubuntu instead of the Amazon Linux environment associated with the previous default runtimes.

Do developers need to migrate existing sandboxes?

Vercel says the previous Sandbox runtimes are deprecated, while new sandboxes created with SDK version 3 default to the managed universal image. The announcement does not provide a final retirement date, so teams should review operating-system-specific scripts and follow Vercel’s documentation before migrating production workflows.