Skip to content

Launch Slicer for the first time on Linux

In this guide you'll start Slicer and launch a microVM on demand by name. Then you'll work inside it, and see how to hand the same workflow over to your coding agent. It takes around five minutes on an x86_64 or Arm64 host.

Each VM is a real Linux machine with its own Kernel and systemd. Docker, Kubernetes and anything else you'd run on a server will just work. A new VM boots in a couple of seconds.

Before you start

Install Slicer and save the license key from your email to ~/.slicer/LICENSE. Then join the slicer group so you don't need sudo for every command.

Give your agent its own computer

Slicer's agent skills teach your coding agent how to launch its own VMs and work inside them, instead of on your machine. They work with Claude Code, Codex, Amp, OpenCode, Cursor and others.

npx skills add slicervm/agent-skills

/plugin marketplace add slicervm/agent-skills
/plugin install use-slicer@slicer

Install slicervm/agent-skills from GitHub

Once Slicer is running, ask your agent for a VM:

Spin up a VM using the slicer skill, clone alexellis/arkade into it and build it.

The agent launches a VM, installs what the build needs, and works inside it. Your machine stays untouched, and the VM is still there for you to slicer shell into afterwards. See Give your agent its own computer for what it runs, or Sandbox your agent in a microVM to run the agent itself inside a VM.

If you haven't started Slicer yet, follow the steps below first. Starting the daemon needs sudo, so it's best to run that one command yourself.

The short version

In one terminal, create a config and start Slicer:

mkdir -p ~/slicer && cd ~/slicer
slicer new vm --count 0 > slicer.yaml
sudo slicer up

In a second terminal, launch a VM and get a shell:

slicer vm add vm --name dev
slicer ready dev
slicer shell dev

Read on for what each step does, and what to try next.

How Slicer is organised

Slicer runs as a daemon on your host. Its default config file is called slicer.yaml, and you can pass a different path to slicer up if you need to.

The config defines one or more host groups. A host group is a template for VMs with the vCPU, RAM, disk, network range and OS image to use.

Most of the time you'll launch VMs on demand by name with slicer vm add. This is how coding agents, sandboxes and dev environments use Slicer. It's what count: 0 is for - Slicer starts with no VMs and you add them as you need them. This guide uses count: 0.

Setting count higher launches that many VMs every time Slicer starts. That's best for appliances and clusters that should always be running, such as a PiHole or the nodes of a Kubernetes cluster.

1. Create a config for the daemon

Create a working directory for Slicer. Your config and each VM's disk image will be stored here.

The first argument to slicer new is the name of the host group, in this case vm:

mkdir -p ~/slicer
cd ~/slicer

slicer new vm --count 0 > slicer.yaml

The host group's name is used in three places:

  • As the prefix for each VM's hostname i.e. vm-1, vm-2
  • When launching a VM into that group i.e. slicer vm add vm
  • To derive each VM's MAC address, so every host group on a machine needs a unique name

We've used vm to keep this guide generic. Feel free to name it after what it's for instead, such as sandbox, agents or k3s.

Take a look at what was generated:

cat slicer.yaml

config:
  host_groups:
    - name: vm
      storage: image
      storage_size: 25G
      persistent: true
      count: 0
      vcpu: 2
      ram_gb: 4
      network:
        bridge: brvm0
        tap_prefix: vm
        gateway: 192.168.137.1/24
  image: "ghcr.io/openfaasltd/slicer-systemd:6.1.90-x86_64-latest"
  hypervisor: firecracker
  api:
    port: 8080
    bind_address: "127.0.0.1"
    auth:
      enabled: true
Field Value Description
count 0 No VMs at start-up. Launch them on demand.
vcpu / ram_gb 2 / 4 The default size of each VM, and the most any one VM in this group can be given. Raise them for bigger VMs.
storage image Each VM's disk is a file in the working directory. It's simple and persistent with no extra setup. See snapshot-based storage for faster clones.
storage_size 25G The size of each VM's disk.
persistent true Keep each VM's disk when it's shut down or Slicer restarts, so it can be relaunched.
network brvm0 VMs join a Linux bridge and get addresses from 192.168.137.0/24.
image slicer-systemd Ubuntu 22.04 with systemd for your host's architecture. Ubuntu 24.04, Rocky Linux and Arch Linux are also available.
hypervisor firecracker cloud-hypervisor and qemu are also supported.
api 127.0.0.1:8080 The API used by the slicer CLI, protected by a token in /var/lib/slicer/auth/token.

Run slicer new --help to see all the options. Every field is covered in the config reference.

2. Start the Slicer daemon

The daemon needs root to manage KVM and networking. It's the only command in this guide that needs sudo.

To try Slicer out, run the daemon in the foreground:

cd ~/slicer
sudo slicer up

slicer up loads slicer.yaml from the current directory. To use another file, pass its path i.e. sudo slicer up ./other.yaml. The OS image is pulled on the first start, which takes a few seconds.

Leave it running and open a new terminal for the next steps. Press Control+C to stop it. If you're connected over SSH, you could start it inside tmux or screen to keep it running after you disconnect.

Or install it as a service from the start

If you already know you want Slicer running permanently, skip the foreground run and install it as a systemd service instead. It'll start on boot and restart if it fails:

cd ~/slicer
sudo slicer service generate --install

This creates a service named slicer that loads slicer.yaml, with ~/slicer as its working directory. Run it from the same directory where you created the config, because that's where Slicer keeps each VM's disk and its state.

Follow its logs with journalctl -u slicer -f. See running in the background for more options.

3. Launch a VM by name

Launch a VM called dev:

slicer vm add vm --name dev

VM created
  Hostname: vm-1
  Name:     dev
  Group:    vm
  Arch:     amd64
  IP:       192.168.137.2/24
  Specs:    2 vCPU, 4GiB RAM, GPUs: 0
  Persistent: true
  Duration: 2.944s
  • vm is the host group to launch into.
  • --name dev gives the VM a friendly name to use with every other command.

slicer vm add returns as soon as the VM is created, whilst it carries on booting in the background. Wait for it to be ready:

slicer ready dev

This returns once the guest agent inside the VM is up, usually in under a second. In scripts, you can combine both steps with slicer vm add vm --name dev --wait.

List your VMs:

slicer ls

AGE  NAME  HOSTNAME  STATUS   IP             RAM   CPU  PERSIST  TAGS
---  ----  --------  ------   --             ---   ---  -------  ----
0s   dev   vm-1      Running  192.168.137.2  4GiB  2    yes

Each VM also gets a generated hostname such as vm-1. Either will work with the CLI, but the name is easier to remember. Add --tag to attach your own metadata i.e. --tag project=website.

4. Work inside the VM

Open a shell:

slicer shell dev
alex@host: ~/slicer
alex@host:~/slicer$ slicer shell dev
Connecting to VM: vm-1
Connected! Press Ctrl+] to exit.
Welcome to Ubuntu 22.04.5 LTS (GNU/Linux 6.1.90 x86_64)

* Image built by SlicerVM (2026)
* Website: https://slicervm.com
* Docs: https://docs.slicervm.com

ubuntu@vm-1:~$ uname -r && uptime
6.1.90
 19:20:57 up 4 min,  0 users,  load average: 0.00, 0.00, 0.00
ubuntu@vm-1:~$

Type exit or press Control+] to leave the shell.

If you leave off the VM name, slicer shell, slicer vm exec and slicer vm logs use the most recently launched VM.

You'll be logged in as the ubuntu user with passwordless sudo. On the Rocky Linux and Arch images the user is slicer. The shell runs over Slicer's guest agent, so no SSH keys or network routes are needed.

Run a one-off command:

slicer vm exec dev -- "uname -a && df -h /"

Linux vm-1 6.1.90 #1 SMP PREEMPT_DYNAMIC Sun Sep 13 00:30:58 UTC 2026 x86_64 x86_64 x86_64 GNU/Linux
Filesystem      Size  Used Avail Use% Mounted on
/dev/root        25G  756M   23G   4% /

Copy files in and out, and add -r for a folder:

slicer cp ./notes.txt dev:~/notes.txt
slicer cp -r ./project dev:~/project
slicer cp dev:~/notes.txt ./notes.txt

Install a web server:

slicer vm exec dev -- "sudo apt-get update -qq && sudo apt-get install -y nginx"

With bridge networking, each VM's IP is reachable directly from the Slicer host:

curl -s http://192.168.137.2 | head -n 4

<!DOCTYPE html>
<html>
<head>
<title>Welcome to nginx!</title>

You can also forward a port to localhost. That's useful when the VM is on an isolated network, or when you're using Slicer from another machine:

slicer vm forward dev -L 3000:127.0.0.1:80

Then open http://127.0.0.1:3000. The forward runs until you press Control+C. Slicer's API already uses port 8080 on the host, so pick a different local port.

5. Launch more, and clean up

Launch as many VMs as your host can hold. Each one can be smaller than the host group's vcpu and ram_gb but not bigger:

slicer vm add vm --name build --cpus 1 --ram-gb 2

VM created
  Hostname: vm-2
  Name:     build
  Group:    vm
  Arch:     amd64
  IP:       192.168.137.3/24
  Specs:    1 vCPU, 2GiB RAM, GPUs: 0
  Persistent: true
  Duration: 1.076s

Add --rm for a throwaway VM whose disk is discarded when it stops.

View a VM's boot log and then delete it:

slicer vm logs build | tail -n 3

[    0.698393] vmmeter[597]: 2026/10/12 19:21:55 Collecting metrics for: vm-2 on port: 34723
[    0.700387] configure-dns.sh[565]: + '[' -f /runner/userdata.sh ']'
[    0.701032] vmmeter[597]: 2026/10/12 19:21:55 VSOCK listening on port: 515

slicer vm delete build

Deleted vm-2 (group vm)

VMs are persistent by default. When you stop and start Slicer, dev will be relaunched with its disk intact along with any files and packages you added.

When Slicer stops, it shuts each VM down cleanly so nothing is lost. For throwaway sandboxes, you can generate the config with slicer new --graceful-shutdown=false so Slicer exits straight away instead.

6. Hand over to your coding agent

With the agent skills installed, your agent uses the same commands you've just run. Some things to try next:

Copy this repo into my dev VM, build it there, and bring the binary back to ./bin.

Install Docker in my dev VM and forward its socket, so I can run docker build locally against it.

Use Slicer to launch a new VM called k3s, install K3s in it, and merge the kubeconfig into ~/.kube/config.

To run a coding agent itself inside a VM, see Sandbox your agent in a microVM.

Optional: SSH into VMs

You don't need SSH to work with Slicer. slicer shell, exec, cp and forward all go through Slicer's API and the guest agent inside each VM.

SSH is handy for tools that expect it, such as scp, rsync or VS Code's Remote SSH. VMs run an OpenSSH server. To add your keys, pass one of these to slicer new:

  • --find-ssh-keys - add the public keys found in ~/.ssh/
  • --github USERNAME - import the keys from your GitHub profile
  • --ssh-key "ssh-ed25519 ..." - add a key directly

They're written to the config under ssh_keys or github_user.

From the Slicer host, connect using the VM's IP:

ssh ubuntu@192.168.137.2

To reach VMs from another machine on your network, print the route commands with slicer vm route ./slicer.yaml and run them on that machine. Never run them on the Slicer host itself, as it will break its networking.

See SSH access to VMs for more.

Optional: Use Slicer from another machine

The slicer CLI is a client for the daemon's REST API. The API binds to 127.0.0.1 by default, but if you expose it over HTTPS, you can drive Slicer from your laptop or anywhere else.

Copy the token from /var/lib/slicer/auth/token on the Slicer host, then point the CLI at the API:

export SLICER_URL=https://slicer.example.com
export SLICER_TOKEN_FILE=~/.slicer/token

slicer ls
slicer shell dev

Everything in this guide works the same way remotely, including cp and forward. See the API reference for exposing it safely with an inlets tunnel or Caddy.

Troubleshooting

Problem Fix
No license found on start-up Your license key is emailed to you when you start a trial or subscribe. Save it to ~/.slicer/LICENSE.
permission denied on /var/lib/slicer/auth/token Join the slicer group with sudo usermod -aG slicer $USER. Then log out and back in or run newgrp slicer.
requested RAM ... exceeds host group limit --cpus and --ram-gb can't exceed the host group's vcpu and ram_gb. Raise them in slicer.yaml and restart Slicer.
Bridge conflict: ... already in use A previous run didn't exit cleanly. Start with sudo slicer up --force.
Your LAN already uses 192.168.137.0/24 Regenerate the config with a free range, i.e. --cidr 192.168.140.0/24.
connection refused from the CLI Slicer isn't running. Check the terminal where you ran slicer up, or systemctl status slicer.

Next steps