Kalisto/ documentation source (permission required) ↗
On this page

#Install and start a cluster

Use the standalone Kalisto binary from any directory. A source checkout and cargo run are not installation prerequisites.

#Install the CLI

Choose your operating system and architecture in the Kalisto download selector. The available packages are development candidates, not a stable release. Download the matching package and its checksum inventory, verify it before installing, and inspect its BUILD-INFO.json source/target metadata. See binary verification for the four supported package targets and platform caveats. Do not replace an existing installation without authorization.

#macOS host prerequisite

The default Apple-container backend requires Apple silicon and macOS 26 or newer, with Apple's compatible container CLI on PATH and its system service already running. The validated runtime version is 1.7.1. For a new Homebrew installation:

brew install container
container system start
container --version
container list

container list must succeed, even with no containers. Skip installation and start if a compatible CLI/service already exists. Never replace, stop or reconfigure a shared running daemon. Runtime/kernel setup happens before the startup measurement. Read Apple prerequisites and firewall requirements before proceeding. Intel Macs and older macOS versions must explicitly choose Lima.

#Prepare and start on macOS

kalisto pull --addons cni/cilium
kalisto up --workers 2 --addons cni/cilium
kalisto kubectl -- get nodes

This creates one server and two workers. Preparation can take several minutes. It is per artifact profile—architecture, K3s/Cilium versions and enabled components—not per cluster or working directory. Repeat pull only to verify/reuse the cache, recover pruned assets, or prepare changed inputs. It does not create cluster state or Kubernetes nodes.

Select Cilium in both commands. Bare pull and up do not enable a CNI. up requires a complete matching cache and does not fetch missing startup assets. Ingress is not installed by default. See offline preparation for supported startup add-ons and private per-cluster TLS manifests.

#Linux Firecracker quick start

Install the Linux host prerequisites, including Firecracker, accessible KVM and rootless slirp4netns networking. Native validation used Fedora 44 x86_64 and Firecracker 1.18.0.

Provide Firecracker-compatible local kernel/rootfs files or explicitly configure manifest/artifact URLs. There is no implicit public artifact source. A Lima disk/template is not a Firecracker rootfs.

KERNEL=/absolute/path/to/kalisto-vmlinux
ROOTFS=/absolute/path/to/rootfs.raw

kalisto pull --backend firecracker --addons cni/cilium \
  --firecracker-kernel "$KERNEL" --firecracker-rootfs "$ROOTFS"
kalisto up --backend firecracker --name dev --workers 2 --addons cni/cilium \
  --firecracker-kernel "$KERNEL" --firecracker-rootfs "$ROOTFS"
kalisto kubectl --name dev -- get nodes

Preparation installs managed Linux Crane, Helm and kubectl and builds a shared read-only ext4 boot-asset disk. No host Docker or manually installed Helm/Crane is needed. up creates fresh VMs, imports local images and applies a private per-cluster Cilium manifest. Missing preparation is an error, not an online fallback. Rootless networking needs no host sudo; --network bridge is an explicit privileged alternative.

#Use and retain your cluster

kalisto ls
kalisto status --name dev
kalisto apply --name dev --manifest app.yaml
kalisto kubectl --name dev -- get nodes
kalisto down --name dev
kalisto up --name dev

Use the actual cluster name (kalisto by default, dev in the Linux example). app.yaml is your manifest; new workload images may need registry access after startup. down retains disks/state and Firecracker SSH identities; up resumes the saved profile without repeating artifact flags. Concurrent clusters need distinct names and free API ports; the default is 16443.

kalisto destroy --name dev irreversibly deletes that named cluster, but leaves prepared artifacts cached. Never use bulk deletion without explicit authorization. Continue with CLI lifecycle, workloads, archives or agent workflows.

#What the startup claim covers

The under-sixty-second acceptance clock includes fresh three-node creation, image imports, authenticated host API, all Cilium rollouts, both-worker DNS and cross-node Service HTTP. Preparation is excluded. Measurements are not universal hardware guarantees or first-install timing promises. See complete results and live-test boundaries.