LogoFreestyle

About Freestyle VMs

An introduction to Freestyle's Virtual Machine (VM) service and what makes it different.

This document provides a high-level overview of Freestyle's VM service, its features, and why to use it.

To get started with Freestyle's VM service, check out the Getting Started Guide.

Overview

Freestyle VMs are full Linux virtual machines designed for speed and flexibility. They start in under a second, pause and resume instantly, and can be forked mid-execution without interruption.

Features

Sub-Second Startup

VMs provision in under 800ms from API request to running machine. Traditional cloud VMs take minutes. This speed comes from memory snapshots—when you get a VM, it's already booted and running, not starting from a powered-off state.

Pause and Resume

VMs can be paused in their running state and resumed later in under 100ms, returning to the exact same memory state. A game server can pause when no players are connected and resume instantly when someone joins—no boot time, no lost state.

Forking

A running VM can be forked without noticeably pausing the original. Fork it N times with minimal performance impact. If an AI agent is browsing a website and wants to explore 20 different paths, fork the VM 20 times and let each copy explore independently—all from the same browser state.

Templates and Caching

VMs are built from layered templates. You can create custom layers, and the system automatically caches them for fast startup. See Templates and Snapshots for details.

Integrations

Integrations add common functionality to VMs: language runtimes like Node.js, Python, Bun, and Ruby, plus databases, browsers, and more. Each integration includes caching for fast startup and a client SDK for easy use. You can also build custom integrations.

Access Control

VMs support multiple Linux users and groups, SSH access, and API access via identities and tokens.

Beyond Sandboxes

Many sandbox platforms restrict what you can run to keep things simple. Freestyle VMs are full Linux environments with low-level access: SSH, systemd, multiple users and groups, and configurable networking. Some sandboxes limit you to building around containers, Freestyle lets you run any Linux-compatible software, including any container, or multiple containers in one VM. Other sandboxes also build out the utilities like runCode as a core part of their platform, which limits what their users can build. Freestyle's flexible integration scheme means our base images are minimal, and you can add exactly what you need. You can see this in how Freestyle has integrations for tons of programming languages, while other sandboxing platforms often support just TypeScript or Python.

Beyond EC2s and Traditional VMs

Traditional cloud VMs are slow — they take minutes to start. Kubernetes on AWS/GCP can take hours to get a new node. They are built around large chunks of compute without secure virtualized isolation of small pieces in mind. While this makes sense for big applications, if you're trying to run lots of small VMs for AI workloads, the overhead is significant and the performance is poor.

On this page

Freestyle AI

Documentation assistant

Experimental: AI responses may not always be accurate—please verify important details with the official documentation.

How can I help?

Ask me about Freestyle while you browse the docs.