How it works
A microVM is still a virtual machine: the Linux kernel on the host (KVM) runs it, and the processor keeps it apart from its neighbours. What changes is the program that builds it. A microVM monitor creates only the devices the guest needs: a network card, a disk, a channel to the host and a serial console.
The best-known monitor is Firecracker. Amazon Web Services built it for AWS Lambda and released it as open source. Its documentation lists three numbers: less than 5 MiB per microVM of overhead, a start in under 125 ms, and up to 150 microVMs per second per host. It is written in Rust and licensed under the Apache License 2.0.
There is no firmware step and no hardware probing. Firecracker loads a Linux kernel directly and hands over control. Your programs run on that kernel, and root inside it is real root.
Why it can sleep and wake
A small device model has one more benefit. The whole state of the machine is small and well defined: memory, processor registers and a handful of devices. That state can be written to disk and loaded back.
Uno does this when your computer has had nothing to do for 10 minutes. Memory and running programs are saved, so an open terminal or a half-finished session is still there when a request wakes the machine. In our measurement on 3 October 2026, restoring the memory took a median 12 ms on the node.
$ sleep 3600 & [1] 4127 … the computer sleeps, then wakes on a request … $ jobs [1]+ Running sleep 3600 &
microVM, VM and container
| Classic VM | microVM | Container | |
|---|---|---|---|
| Own kernel | Yes | Yes | No, shares the host's |
| Virtual devices | Many, like a PC | A handful | None, it is a process |
| Starts | Like a PC booting | In a fraction of a second | In a fraction of a second |
| Can be saved with memory | Yes | Yes, and it is small | Rarely |
| Runs | Almost any OS | Linux | Linux programs |
You can think of a microVM as a container's speed with a VM's wall. Our comparison page covers VPS, VM, container and sandbox side by side.
Do you need one?
Take a microVM when
- You start many machines. Each one is cheap to create and to keep.
- Code runs that you did not write. An AI agent, a customer's script. Each gets its own kernel.
- Machines idle most of the time. A saved microVM costs disk space, not memory.
Skip it when
- You need a desktop or a screen. A microVM has no graphics. You reach it over the network.
- You need another operating system. Firecracker microVMs run Linux.
- You run one app on your own server. A plain container is simpler.
A microVM in Uno
Every Uno computer is Firecracker microVM, one per computer, root inside. You do not run Firecracker yourself and you do not build images: you get a Linux machine with Docker and your own disk, and Uno handles sleep and wake. How the computer works.
Small: $5 a month
One microVM per computer
Every plan is the same kind of machine. Bigger plans get more processors and memory. Move up in the console and your files come with you.
Small
A Telegram bot, a VPN, a small backend
$5/ month
- 1 vCPU · 1 GB
- 15 GB disk
- 25 GB cloud storage
- 5 boosts a month
Plus
Agents that work with your laptop closed
$20/ month
- 4 vCPU · 4 GB
- 40 GB disk
- 100 GB cloud storage
- Uno Work in the cloud
Pro
A product in production, a second MacBook
$70/ month
- 16 vCPU · 16 GB
- 120 GB disk
- 300 GB cloud storage
- Speed of a MacBook Air M4
Max
A team and many agents at once
$200/ month
- 32 vCPU · 48 GB
- 300 GB disk
- 1,024 GB cloud storage
- 5 public IPv4
Add Uno AI to any plan: 40 h of AI work a month on Small, 100 h on Plus, unlimited on Max. Today: GLM-5.3 Flash and MiMo v2.6 Pro. Or sign in with your own Claude or ChatGPT.