Vercel Sandbox now runs on Vercel Managed Images

Vercel introduced Vercel Managed Images (VMI), a set of versioned, open-source base images that replace the deprecated Sandbox runtimes, with the new default `vercel/sandbox/universal:latest` shipping Node.js 24, Python 3.14 with `uv`, and coding agents including `opencode`, `claude-code`, `codex`, and `pi` on Ubuntu 26.04. The images receive nightly releases with rolling tags that auto-update dependencies, while digest-pinned images offer immutable environments. Managed images run as the default `ubuntu` or `arch` user with passwordless sudo, and the previous `runtime` property remains deprecated but functional for teams needing Amazon Linux.

Today we are introducing Vercel Managed Images VMI , a set of versioned, open-source base images you can use as-is or extend. The source for every image lives in the public vercel/sandbox https://github.com/vercel/sandbox repository. Managed images replace Sandbox runtimes, which are now deprecated. Starting with version 3 of the Sandbox SDK, new sandboxes default to vercel/sandbox/universal:latest . It ships with Node.js, Python, common coding agents and standard utilities; so most users never build a custom image or install packages at boot. The universal image switches our default operating system from Amazon Linux to Ubuntu, a lighter and more widely used system across the industry. Every managed image gets a nightly release. Rolling tags such as latest and the major-version tags pick up operating system and dependency updates automatically, including security patches and new releases of Node.js, Python, and the preinstalled coding agents. Within each release, dependencies are pinned to specific versions wherever possible, so a given image version stays consistent. If you need a fully immutable, reproducible environment, pin to an image digest SHA . Digest-pinned images opt out of automatic updates. Each managed image targets a different starting point: vercel/sandbox/universal:latest is the new default, a rolling release on Ubuntu 26.04 with Node.js 24, Python 3.14 with uv , the opencode , claude-code , codex , and pi coding agents, and utilities including git , vim , nano , tmux , ripgrep , jq , and fzf . vercel/sandbox/node:22 , node:24 , and node:26 provide Node.js versions on Ubuntu 26.04. vercel/sandbox/python:3.14 provides Python on Ubuntu 26.04. vercel/sandbox/ubuntu:latest provides base Ubuntu 26.04 with no extra tooling. vercel/sandbox/arch:latest provides Arch Linux, regularly updated, with no Node.js or Python preinstalled. Its large package repository makes it useful for agents that install tools on the fly. Sandboxes can use an image from any repository in your project, a repository which has been shared with your team https://vercel.com/changelog/share-vercel-container-registry-repositories-across-teams or any public repository https://vercel.com/changelog/vercel-container-registry-repositories-can-now-be-made-public . Images are addressed by their full path, team-slug/project/repo:tag , for example vercel/sandbox/universal:latest . The SDK types the image property so built-in images autocomplete while any string is still accepted. The previous runtime property is deprecated, not removed, so existing code keeps working. Amazon Linux runtimes are not part of the managed image catalog, so teams that need AL2023 can stay on runtime . Managed images run as the default ubuntu or arch user with passwordless sudo, rather than the vercel-sandbox user. Open an issue or a pull request to improve them, and read the Vercel Sandbox images documentation https://vercel.com/docs/vercel-sandbox/concepts/images to get started.