cloudflavor / kalisto

Local Kubernetes.
Your k3s. Your machine.

A standalone CLI for local k3s clusters. Apple containers are the macOS default. Lima is optional; Linux Firecracker supports prepared offline Cilium startup with rootless networking.

01 / distribution

Download & install

Development snapshot 0.0.0-ci.4476+g015c15f4bb23 (not a stable release), from 015c15f4bb23. See verification and provenance.

Source links point to the canonical private GitHub repository and require an authorized account. Downloads and checksums on this site are public; repository access is not required to download them.

Download .tar.gz ↓

kalisto-ci-015c15f4bb23-macos-arm64.tar.gz

SHA256SUMS ↗ · Each package contains kalisto and its BUILD-INFO.json provenance.

macOS binaries are unsigned and not notarized. Verify the checksum and source before deciding whether to allow the executable in macOS security settings. Do not disable Gatekeeper.

Host prerequisite: Apple silicon, macOS 26+, and Apple container installed through Homebrew with its system service running. Reuse a compatible installation; do not replace or stop a shared daemon. Follow the macOS prerequisite guide before starting Kalisto.

Or download, verify, and install into your user-owned executable directory:

shell / macOS · ARM64
curl -fLO 'https://cloudflavor.io/downloads/kalisto/ci-015c15f4bb23/kalisto-ci-015c15f4bb23-macos-arm64.tar.gz'
printf '%s  %s\n' '01f68d705848c96ab16fc1b8d7a7578c653218e0f58058343955ef877507beab' 'kalisto-ci-015c15f4bb23-macos-arm64.tar.gz' | shasum -a 256 -c -
mkdir -p "$HOME/.local/bin"
tar -xzf 'kalisto-ci-015c15f4bb23-macos-arm64.tar.gz' -C "$HOME/.local/bin" kalisto
export PATH="$HOME/.local/bin:$PATH"

Keep ~/.local/bin on your shell’s PATH. Checksums detect corruption; a checksum from the same host is not independent publisher authentication.

02 / choose your backend

Prepare once. Start a functional cluster.

Prepare matching assets before startup; up does not download them. CNI defaults to none; choose Cilium explicitly for pod networking. The getting-started guide covers installation and a fresh three-node cluster.

Apple containers macOS default

Apple silicon and macOS 26+, with Homebrew Apple container installed and running. Kalisto manages its own k3s, not Apple’s experimental Kubernetes implementation.

Apple prerequisites and troubleshooting →

Lima explicit alternative

Native macOS VZ for Intel or older Macs. Choose Lima explicitly; Kalisto never silently falls back or converts an existing cluster to another backend.

Optional Lima backend →

Firecracker Linux

Prepared offline Cilium startup with rootless networking, validated on Fedora 44 x86_64: 40.68 seconds preparation, then 42.95 seconds for fresh three-node startup, authenticated API, Cilium, worker DNS and cross-node HTTP probes. Preparation is outside the startup clock.

Supply kernel/rootfs files or configure artifact URLs; no implicit public artifact source is provided.

Linux setup and validation limits →

03 / real terminal use cases

From prepared cluster to everyday work.

Two backend showcases, each following an isolated cluster through the same four real CLI use cases: start, deploy, resume and scale. Watch them in the embedded asciinema.org players below. Playback starts only when you press play.

Full real-time playback: original timing and all waits are preserved at 1× speed. Host installation, service setup and artifact preparation are excluded. These scenes do not promise a fresh install in under 60 seconds or replace the full functional performance acceptance checks.

The commands below are examples, not a script to run blindly. Replace angle-bracket values with your own unused cluster name, available API port and prepared artifact paths. Download the fixture matching your backend and architecture as app.yaml.

Apple container on macOS

macOS · Apple silicon · Captured · kalisto 0.1.0-dev.1+g93390c52e4f0
Source 93390c52e4f0 · Download macOS · Apple silicon fixture ↓

Requires Apple silicon, macOS 26+ and a running Apple container service. Follow the Apple container prerequisite guide before preparing assets.

kalisto pull --backend apple-container --addons cni/cilium

Start a prepared cluster

Start one server and two workers with explicit Cilium networking, then inspect authenticated node readiness. Prepare matching assets first with the backend-specific pull example above and its prerequisite guide.

kalisto up --backend apple-container --name <your-cluster> --api-port <free-api-port> --workers 2 --addons cni/cilium

Prepare and start your cluster →

Loading player from asciinema.org…

Watch on asciinema ↗· Original timing, 1× speed

Recording provenance

Original duration: 34.435992 seconds
SHA-256: 652981327f1fce33e5504e35e26921f6e048508d017efa2e7a565a5f009676b3

Deploy and reach an HTTP app

Apply the public, digest-pinned HTTP fixture from prepared local images, inspect its pods and Service with managed kalisto kubectl, and check a real HTTP response through a loopback port-forward.

kalisto apply --name <your-cluster> --manifest app.yaml

Apply workloads and inspect networking →

Loading player from asciinema.org…

Watch on asciinema ↗· Original timing, 1× speed

Recording provenance

Original duration: 2.583711 seconds
SHA-256: 640babcb08ed329093ae3f9fff3bfea43f5be9e0caadba756918ebe7923114a8

Stop, resume, keep your state

Save a ConfigMap, stop only the demo cluster, then resume the same profile and read back the saved state and HTTP workload. down retains state; destroy deletes it.

kalisto down --name <your-cluster>
kalisto up --backend apple-container --name <your-cluster> --api-port <free-api-port> --workers 2 --addons cni/cilium

Understand stop and resume →

Loading player from asciinema.org…

Watch on asciinema ↗· Original timing, 1× speed

Recording provenance

Original duration: 31.795279 seconds
SHA-256: a3077229b217ca7a21e009175d1a116d364ab1c5bf48055e5201f9ecd0f68989

Add a worker

Grow from two workers to three, verify all four nodes are Ready, then spread three HTTP replicas across distinct workers. The worker count excludes the server.

kalisto scale --name <your-cluster> --workers 3

Scale and manage your cluster →

Loading player from asciinema.org…

Watch on asciinema ↗· Original timing, 1× speed

Recording provenance

Original duration: 23.536375 seconds
SHA-256: 90eaf1bcf681d8a1b46f38dd18ca078a94282773b7c2475d4e2bf94367014851

Firecracker on Linux

Linux · x86_64 · Captured · kalisto 0.1.0-dev.1+g4fe2abc74675
Source 4fe2abc74675 · Download Linux · x86_64 fixture ↓

Prepare Linux KVM, Firecracker, guest-disk and networking tools first. Both pull and up require explicit kernel/rootfs flags for these examples; no turnkey public images are implied. Follow the Firecracker prerequisite and artifact guide for compatible files and preparation.

kalisto pull --backend firecracker --firecracker-kernel <kernel-path> --firecracker-rootfs <rootfs-path> --addons cni/cilium

Start a prepared cluster

Start one server and two workers with explicit Cilium networking, then inspect authenticated node readiness. Prepare matching assets first with the backend-specific pull example above and its prerequisite guide.

kalisto up --backend firecracker --firecracker-kernel <kernel-path> --firecracker-rootfs <rootfs-path> --name <your-cluster> --api-port <free-api-port> --workers 2 --addons cni/cilium

Prepare and start your cluster →

Loading player from asciinema.org…

Watch on asciinema ↗· Original timing, 1× speed

Recording provenance

Original duration: 44.488632 seconds
SHA-256: 368e5e628331d528fb5e19bd8cbe4bbdf4534edb22ef89b8c321240c2574df21

Deploy and reach an HTTP app

Apply the public, digest-pinned HTTP fixture from prepared local images, inspect its pods and Service with managed kalisto kubectl, and check a real HTTP response through a loopback port-forward.

kalisto apply --name <your-cluster> --manifest app.yaml

Apply workloads and inspect networking →

Loading player from asciinema.org…

Watch on asciinema ↗· Original timing, 1× speed

Recording provenance

Original duration: 1.935188 seconds
SHA-256: c484bf7ca2c041140c4cc1eec25fca2c48f72fd10758bf92f86753d98944e51a

Stop, resume, keep your state

Save a ConfigMap, stop only the demo cluster, then resume the same profile and read back the saved state and HTTP workload. down retains state; destroy deletes it.

kalisto down --name <your-cluster>
kalisto up --backend firecracker --firecracker-kernel <kernel-path> --firecracker-rootfs <rootfs-path> --name <your-cluster> --api-port <free-api-port> --workers 2 --addons cni/cilium

Understand stop and resume →

Loading player from asciinema.org…

Watch on asciinema ↗· Original timing, 1× speed

Recording provenance

Original duration: 47.639932 seconds
SHA-256: a95332275cf9a7ccc0fa6c1645d54edbab51e258661158ae19b549209b3aef94

Add a worker

Grow from two workers to three, verify all four nodes are Ready, then spread three HTTP replicas across distinct workers. The worker count excludes the server.

kalisto scale --name <your-cluster> --workers 3

Scale and manage your cluster →

Loading player from asciinema.org…

Watch on asciinema ↗· Original timing, 1× speed

Recording provenance

Original duration: 30.595535 seconds
SHA-256: 3b615aceec61d296f340b2e3ca1ffc4a5c9cc7f57e765b21d2d0e5090c27b6f3

04 / canonical documentation

One place for every guide.

Installation, preparation, backend troubleshooting, CLI reference, workloads, Lima-only archives, AI agents, performance evidence and maintainer tooling live in the Kalisto documentation.