install + usage ​

# install (macOS + Linux)
# option 1: download from GitHub releases
# https://github.com/smol-machines/smolvm/releases

# option 2: install script
curl -sSL https://smolmachines.com/install.sh | bash

# sandbox - ephemeral isolated environments
smolvm sandbox run --net alpine -- echo "hello"
smolvm sandbox run --net alpine -v ./data:/workspace -- ls /workspace
smolvm sandbox run --net python:3.12-alpine -- python -V

# microvm - persistent linux VMs
smolvm microvm create --net myvm
smolvm microvm start myvm
smolvm microvm exec --name myvm -- apk add git  # changes persist across reboots
smolvm microvm exec --name myvm -it -- /bin/sh   # interactive shell
smolvm microvm stop myvm

# smolfile - declarative VM configuration
smolvm sandbox run -d -s examples/openclaw-app/openclaw.smolfile alpine/openclaw:main -- openclaw gateway --port 18789
smolvm microvm create myvm -s my-app.smolfile

# pack - portable, executable virtual machine
smolvm pack create alpine -o ./my-sandbox
./my-sandbox echo "hello"

smolvm pack create python:3.12-alpine -o ./my-pythonvm
./my-pythonvm python3 -c "import sys; print(sys.version)"

# uninstall
curl -sSL https://smolmachines.com/install.sh | bash -s -- --uninstall

about ​

smolvm runs Linux microVMs on your machine. No daemon, no Docker, no cloud account.

A microVM is a lightweight virtual machine — hardware-level isolation with <200ms boot time. Your host filesystem, network, and credentials are completely separated from the workload unless you explicitly share them in.

Install a single binary, run a command, get an isolated Linux environment.

use this for ​

• sandbox untrusted code (pip install, npm install, AI agents) with hardware isolation
• run Linux microVMs on macOS and Linux — no Docker daemon needed
• declare workloads with a Smolfile and run, pack, or deploy them
• mount host directories for shared file access, with the host protected by a hypervisor boundary
• pack and distribute self-contained sandboxed applications as portable .smolmachine executables

comparison ​

smolvm is designed for running local (on your machine), persistent (long running and stateful), and interactive (communicate with workload/vm in real time) workloads.

[1] container isolation · [2] containerd benchmark · [3] QEMU boot time · [4] Firecracker website · [5] Kata Containers · [6] Kata boot time · [7] Kata installation · [8] Firecracker requires KVM · [9] Kata macOS support

vs colima + krunkit ​

Colima runs one Linux VM that hosts a container daemon (containerd/dockerd). All your containers share that single VM. smolvm starts a separate microVM per workload — each with its own kernel, filesystem, and isolation boundary. A compromised container in colima can reach other containers in the same VM. A compromised workload in smolvm is confined to its own VM with no access to the host or other workloads.

smolvm also ships a custom-built kernel (libkrunfw) optimized for fast boot, and produces portable .smolmachine executables that run without any runtime installed.

how it works ​

libkrun VMM + Hypervisor.framework (macOS) / KVM (Linux) + crun container runtime. No daemon process — the VMM is a library linked into the smolvm binary.

platform support ​

made by @binsquare · twitter · github