Hyperconvergence for home, made simple.

Apps, virtual machines, replicated storage and networking across ordinary machines - in one friendly interface.
Guided Kubernetes setup and integrated updates, designed for your homelab.

Live demo  ·  Quick install  ·  Features  ·  Wiki  ·  Reference

Try it before you install it: the live demo is the real interface on made-up data, in your browser. Click around every page; nothing you do there is saved.

Why Homestead

A homelab on a single machine is one failed disk, power supply or update away from going dark. Start there, then grow across ordinary machines - mini PCs, old desktops, virtual machines - with spare capacity at each layer. Host failover needs control-plane quorum, healthy replicas and eligible spare hosts; three server nodes are the usual starting point for an HA control plane.

Only two machines? Prefer two linked single-node clusters for independent operation. Two embedded-etcd servers need both online; losing either loses quorum. For one HA k3s cluster, add a suitable third voting server, such as an ARM64 Raspberry Pi with SSD storage. Linking clusters provides shared management, not automatic failover.

Compute Eligible containers and VMs can restart on a healthy node when one fails. Placement rules decide where each app may run, and compatible VMs can live-migrate between hosts for maintenance.

Storage Longhorn replicates configured volumes across nodes to protect against disk or host failure. Snapshots, scheduled backups to S3 or NFS, and restores are managed in the same interface.

Network Each app can have its own virtual IP, announced by kube-vip, that stays reachable when a node goes down. IP address management keeps track of every address on your LAN.

It feels like a NAS, not a cluster. Homestead speaks in apps, volumes, shares and VMs - the words you know from Unraid or Docker - and looks after the Kubernetes underneath. When you want the raw objects, every resource is a click away as YAML. It runs on k3s, RKE2, Harvester HCI or any Kubernetes you already run.

Quick install

Choose k3s for home self-hosting. It is Homestead's preferred platform and the most widely tested with Homestead. Start with the k3s guide; RKE2 and Harvester are supported alternatives.

On a Linux machine - a bare one, a k3s or RKE2 server, or a Harvester node:

curl -sfL https://raw.githubusercontent.com/homestead-lab/homestead/main/scripts/install.sh | sudo sh

• New cluster or join one. The installer checks the machine, then creates a k3s or RKE2 cluster - or joins this machine to an existing one - with Longhorn, optionally KubeVirt, and Homestead. The installation summary lets you choose the version of each component.
• More machines. Run the same line on each further machine to join it.
• Node doctor. On any node, the same line checks the node and the cluster and offers a fix for each problem it finds.

Disks on a k3s or RKE2 machine. Give the system 64-128 GB and keep the rest for Longhorn:

• A second drive is best. Leave it blank: Nodes → Disks → Add to Longhorn formats and mounts it safely, or keeps Longhorn data already on it.
• One drive? Install the OS on LVM (Ubuntu Server's default) and leave the volume group's spare space unallocated - Ubuntu gives the root volume 100 GB and leaves the rest free. Use its free space on the system disk turns it into a Longhorn disk of its own, keeping a tenth (at least 10 GB) for the system to grow into. Homestead does not resize the partitions of a running system, so plain partitions without LVM leave nothing it can use.
• Longhorn's V2 engine needs a whole drive of its own in Homestead's guided disk workflow, plus CPU and memory headroom. Read the V2 support and considerations before enabling it, particularly on ARM or mixed-size hosts.

See Installing on k3s.

Then open http://<your machine>:8088 and create the administrator account. Already running a cluster? Install with Helm or the manifest. The wiki walks through k3s, RKE2 and Harvester step by step.

Features

App Store - Unraid's Community Applications catalogue, deployed properly: volumes, ports, hardware and updates handled for you. Containers - deploy, edit, logs and console; groups, autostart and placement; update checks with a monitored rollout and one-click rollback, or automatic updates in your maintenance window - snapshotted first, rolled back if the app stops answering. Logos from the app store, change history with undo, log search across every app, and start and stop schedules. Reservations suggested from a week of real use, set for many apps at once, and priority tiers that say which apps keep running when the cluster is short of room. Homestead updates itself from its own button on the top bar. Monitoring - every app and VM asked every minute whether it answers, with 30 days of history, alerts, a Monitoring page and dashboard widget, and - only if you ask - a restart or a webhook when one is down. Virtual machines - KubeVirt VMs from a store of cloud images, your own disks, or ISOs from your shares; hardware settings with Windows and Linux presets (UEFI, Secure Boot, TPM, CPU model and pinning); a console, power actions, live migration, monitoring and start and stop schedules. Data protection - snapshots, recurring backups, restores, and moving containers and VMs between clusters, and restore tests that prove a backup brings an app back. Browse snapshot files read-only and download them while the live container or SMB volume stays online. Networking - a virtual IP per app, collision-free port exposure, which node answers for each address and whether it is reachable (with kube-vip's missed addresses recorded for it), and IP address management with scanning and UniFi sync. Imports - containers and VMs from an Unraid server, a Docker Compose file, or a VM disk image (qcow2, VMDK). AI agents - an MCP server so Claude Code and other agents can look after the cluster, with a person approving anything destructive (guide). API - scoped, expiring API keys for Home Assistant, scripts and AI agents, on a versioned, self-describing API (OpenAPI, guide).

Nodes and hardware - node health, SMART drive health, disks, iGPU, Coral and USB passthrough, and a terminal on every node. One placement choice for the whole cluster - spread by capacity or a hot spare - and containers moved back when a host returns. Cluster - control-plane and etcd health, adding and removing hosts, and platform versions: k3s or RKE2, Longhorn, KubeVirt, CDI, kube-vip and Multus upgraded a step at a time, Harvester upgrades followed. Dashboard and alerts - 90 days of history, push notifications, and MQTT with Home Assistant discovery; a forecast of when storage fills, and what would stop or move if a host went down, and whether the other hosts have room for it. Network shares and Portal - Samba shares from any volume, and a tile for every web interface on your network. Helm and Resources - every Helm release and every Kubernetes object, with its YAML and events. Linked clusters - several clusters from one sign-in: switch between them, see their containers and VMs together, and move workloads across. Safe by design - viewer, operator and admin roles, guarded deletion, and every job cancellable, with a rollback.

How Homestead compares

Hyperconvergence for home, with apps and infrastructure in one friendly interface. Choose K3s (recommended), RKE2 or Harvester underneath. Start with one small machine and grow into a cluster, on supported x86-64 or ARM64 hardware.

✅ Built-in or guided · ⚪ Supported with add-ons or custom configuration · ❌ Not supported by the product

Feature	Homestead	Unraid	Proxmox VE	CasaOS	Harvester	Talos Linux	MicroCloud	Headlamp
Web management	✅	✅	✅	✅	✅	⚪	✅	✅
Mobile app / PWA	✅	⚪	✅	❌	❌	❌	❌	❌
Alerts and notifications	✅	✅	✅	⚪	✅	⚪	⚪	⚪
Self-hosted app store	✅	✅	⚪	✅	⚪	⚪	⚪	✅
Helm chart management (UI)	✅	⚪	⚪	⚪	⚪	⚪	⚪	✅
Application containers (Docker / OCI images)	✅	✅	✅	✅	⚪	✅	⚪	✅
Linux system containers (LXC / LXD)	❌	⚪	✅	❌	❌	❌	✅	❌
Single sign-on (SSO)	❌	✅	✅	❌	⚪	⚪	✅	✅
Virtual machines	✅	✅	✅	❌	✅	⚪	✅	⚪
GPU / PCI passthrough to VMs	✅	✅	✅	❌	✅	⚪	✅	⚪
USB devices in containers	✅	✅	✅	⚪	⚪	⚪	✅	⚪
Single host → multi-node cluster	✅	❌	✅	❌	✅	✅	✅	⚪
ARM64 hosts	✅	❌	✅	✅	✅	✅	✅	✅
Distributed, replicated storage	✅	❌	✅	❌	✅	⚪	✅	⚪
Automatic workload failover	✅	❌	✅	❌	✅	⚪	✅	⚪
IP address management (IPAM)	✅	⚪	✅	⚪	✅	⚪	✅	⚪
Network file shares	✅	✅	⚪	✅	⚪	⚪	⚪	⚪
Workload snapshots and backups	✅	⚪	✅	⚪	✅	⚪	✅	⚪
Home Assistant / MQTT monitoring	✅	⚪	✅	⚪	⚪	⚪	⚪	⚪
Guided setup and upgrades	✅	✅	✅	✅	✅	⚪	✅	⚪
App uptime checks and alerts	✅	⚪	⚪	⚪	⚪	⚪	⚪	⚪
Automatic app updates with rollback	✅	⚪	⚪	⚪	⚪	⚪	⚪	⚪
Scheduled start and stop	✅	⚪	⚪	⚪	⚪	⚪	⚪	⚪

Helm means managing charts through a UI. Headlamp's ticks include its official App Catalog; infrastructure features depend on the connected Kubernetes cluster. Homestead's Helm changes require the K3s/RKE2 Helm controller. Application containers package apps or services such as Jellyfin, Pi-hole or MQTT. Homestead runs Docker / OCI images through Kubernetes. LXC/LXD system containers provide a fuller Linux environment, like a lightweight server sharing the host's kernel. SSO means identity-provider login to management; an external access gate alone does not count. The last three rows mean built into the management interface: elsewhere they take a separate monitor (Uptime Kuma, say), an update tool (Watchtower, Renovate) or scheduled scripts.

Compared with Proxmox: Homestead brings an app store, hardware-aware app placement, managed file shares, LAN address inventory and MQTT discovery together with cluster administration. Its phone interface is an installable PWA with push notifications. Proxmox also provides passthrough, distributed storage, backups, IPAM, alerts and mobile access.

Homestead is beta. Hardware, architecture and HA requirements vary by stack; device-bound workloads restrict migration and failover. Large-scale deployments have not been benchmarked. Reviewed 3 October 2026, with the last three rows added 7 October; details, qualifications and sources.

On your phone

Homestead installs as an app on iOS and Android (over HTTPS), with push notifications for outages, degraded storage, failed jobs and image updates.

On the desktop

Containers	Architecture
Volumes	Virtual machines

The live demo opens with a healthy cluster. For warning and failure examples, use ?demo-scenario=incidents or ?demo-scenario=critical on the demo URL. Local demos add these to ?demo=1. Page audits check the healthy default, dialog audits use the incident scenario, and CI checks all three health states.

Every screenshot is taken from Homestead's demo data by each release, so they always show the current version. The same demo runs live at homestead-lab.github.io/homestead, updated with each release.

Platforms

k3s is preferred for home self-hosting and is the most widely tested platform with Homestead. RKE2 and Harvester remain supported alternatives. Homestead checks what the cluster has, and each page works with that.

	k3s (recommended)	Harvester	RKE2 or other Kubernetes
Install	one line from bare Linux	one line on a node, or Helm or the manifest	RKE2: one line from bare Linux, or onto an RKE2 server; any other: Helm or the manifest
Containers, App Store, Compose, Portal, Networking, IP addresses, Resources, dashboard	yes	yes	yes
Volumes, data protection, disks	yes, with Longhorn - the installer adds it, or Settings → Hardware and storage → Add-ons	yes, Longhorn is built in	yes, with Longhorn - Settings → Hardware and storage → Add-ons installs it on RKE2
Virtual machines	yes, with KubeVirt - the installer adds it, or Add-ons	built in	yes, with KubeVirt - Add-ons installs it on RKE2
Addresses for apps	a VIP per app with kube-vip, and LAN addresses with Multus (and macvtap for VMs, or a host bridge Homestead can make) - installed by Homestead - beside the nodes' own addresses (ServiceLB)	a VIP per app (kube-vip), LAN addresses (Multus)	RKE2 from the installer: as k3s; others: MetalLB or kube-vip
Helm charts	yes (k3s's Helm controller)	yes (RKE2's Helm controller)	yes on RKE2; listing only without a Helm controller
Adding a host	the join command for a worker or a server	a guide to Harvester's installer	RKE2's join commands
Platform upgrades	k3s, Longhorn, KubeVirt, CDI, kube-vip, Multus and macvtap, a step at a time	followed on the Cluster page	RKE2: as k3s; others: shown

Documentation

• Wiki - the guided tour: building a cluster, installing Homestead, then each page in turn.
• Reference - every feature in depth, security notes, and how releases are made.
• Troubleshooting - the problems people actually hit, and the node doctor.

Contributing

See CONTRIBUTING.md for how the project is put together, running it locally against demo data, the tests, and how releases are made.

Supporters and featured on

Homestead is listed in the HomeLab Addiction self-hosted directory. Thanks for supporting the project.

Licence

Homestead is source-available under the Apache License 2.0 with the Commons Clause 1.0. You can use it at home or internally at work, study the code, modify it, contribute, and redistribute it subject to those terms. Selling Homestead, or a product or service whose value derives entirely or substantially from its functionality, requires separate permission. This can include paid hosting and related consulting or support; the full licence defines the restriction.

Community contributions are welcome. This is a source-available licence, rather than an OSI-approved open-source licence. Previously published MIT versions keep their MIT permissions. See the licensing guide for examples and the transition details.

Third-party libraries, images and catalogue content retain their own licences; Homestead's Commons Clause does not restrict their rights. Their terms and attributions are listed in THIRD_PARTY_NOTICES.md.

Unraid® is a registered trademark of Lime Technology, Inc. Homestead is not affiliated with, endorsed, or sponsored by Lime Technology, Inc.