Your servers.
Your control plane.
Provision VMs and microVMs, automate ZFS snapshots and replication, and run a cluster from the CLI or web UI — on hardware you own.
hoster-lap-01 Quick actions
hoster0103 Quick actions hoster0104 Quick actions hoster0105 Quick actions hoster0106 Quick actions
QEMU / KVM
Cloud Hypervisor
OpenZFS
WireGuard
LeilFS
Prometheus
Ubuntu
Rocky Linux
GoEvery host runs itself. Central visibility when you want it.
Each server manages its own VMs, snapshots and replication. Add Control Center when you need a shared API and web UI across hosts — the servers keep running without it.
hoster0103hoster0104hoster0105hoster0106Control Center is off, so the shared web UI and API go with it. Every host still runs its own VMs, snapshots and replication, your terminal still reaches each one over SSH, and the mesh keeps routing between them.
What the Control Center actually does.Every panel below is the real interface, reproduced from a live five-host fleet. Pick a page to see it.
Fleet
Workloads
Operations
The whole estate on one screen
Totals across the fleet — hosts, allocated vCPU, memory, pool capacity and snapshot space — over tables listing every host, VM, microVM and backup. A problems tab collects everything currently wrong.
- Per-host CPU, RAM, power, pool usage and uptime
- Backup coverage per machine, with remote holders
- Platform and pending-reboot state per host
| Host | CPU | RAM | Power | rpool | VMs | Uptime |
|---|---|---|---|---|---|---|
| hoster-lap-01 | 3.4% | 61.2% | 43.0 W | 28.9% | 5 / 5 | 18d 17h |
| hoster0103 | 3.4% | 36.2% | 6.4 W | 55.8% | 2 / 3 | 55d 1h |
| hoster0104 | 3.2% | 25.8% | 7.6 W | 21.0% | 1 / 1 | 118d 3h |
| hoster0105 | 7.2% | 46.0% | 12.1 W | 48.7% | 3 / 3 | 12d 15h |
| hoster0106 | 6.7% | 31.8% | 4.4 W | 31.1% | 2 / 3 | 131d 14h |
The fleet, and one host in detail
Every host in one list with its state, alert count and load average. Opening one gives you the whole machine: configuration, live CPU, memory and power draw, ZFS pool usage, and every systemd unit with start, stop, restart and logs.
- Tags, maintenance flags and pending-reboot warnings
- Storage pools with capacity and health
- Serial and SSH consoles in the browser
- Markdown notes per host
Machines, their guests and their copiesMicroVMs get exactly the same pages — list, detail, snapshots and monitoring — so there is nothing new to learn for container workloads.
The VM list carries the guest operating system, run state and how many remote replicas exist. A machine’s own page adds its parent host, vCPU topology, memory, network, disk and dataset — plus snapshots, monitoring and a console.
- Full KVM/QEMU machines, with a console and live migration
- Cloud Hypervisor microVMs for Docker and Podman
- PCI and GPU passthrough for training and inference
- Any guest: Debian, Ubuntu, AlmaLinux, Fedora, Rocky, Windows
grafanaProd Debian 12
live-migration-test AlmaLinux 10
webos-run-package-01 Fedora 44 hosteros-lab-0104 Debian 13Every version of a machine, and the way back
Snapshots arrive on a schedule and are kept by retention class. Any row can be cloned into a new machine or rolled back in place, and the page totals exactly how much space the history is costing.
- Retention by tag — snap_frequent through snap_yearly
- Clone or roll back from the row it sits on
- Incremental replication to another host, counted per VM
- Encrypted datasets — rpool/hoster/vm-encrypted
| # | Snapshot name | Date created | Type | Size | Actions |
|---|---|---|---|---|---|
| 1 | hourly_20260803_100010.625877 | 2026-08-03 10:00:10 | hourly | 2M | CloneRollbackDelete |
| 2 | hourly_20260803_090010.623936 | 2026-08-03 09:00:10 | hourly | 4M | CloneRollbackDelete |
| 3 | hourly_20260803_080010.653798 | 2026-08-03 08:00:10 | hourly | 4M | CloneRollbackDelete |
| 11 | daily_20260803_000011.197775 | 2026-08-03 00:00:11 | daily | 324K | CloneRollbackDelete |
| 13 | hourly_20260802_230010.590129 | 2026-08-02 23:00:10 | hourly | 4M | CloneRollbackDelete |
The automation behind the snapshots
Per host, a scheduler registers snapshot, replication, disaster-recovery and template jobs. Defaults can be generated in one click, every job takes ordinary cron syntax, and upcoming runs are listed with the time remaining.
- Snapshot classes from every fifteen minutes to yearly
- Replication jobs with remote host and rate limit
- Template sync for VMs and microVMs, local and remote
- DR schedules generated from each machine’s own config
| Snapshot schedule | Cron | Tag | Keep | Enabled |
|---|---|---|---|---|
| Frequent default | */15 * * * * | snap_freq | 15 | Yes |
| Hourly default | @hourly | snap_hourly | 24 | Yes |
| Daily default | @daily | snap_daily | 7 | Yes |
| Weekly default | @weekly | snap_weekly | 4 | Yes |
| Monthly default | @monthly | snap_monthly | 12 | Yes |
| Yearly default | @yearly | snap_yearly | 2 | Yes |
| Job | Type | Cron | Next run | Targets |
|---|---|---|---|---|
| replicate_vaultwarden-prod | Replication | 13 * * * * | in 2m | hoster0105 |
| template-replication-hourly | VM Remote Sync | 43 * * * * | in 32m | 85 MB/s limit |
| microvm-template-local-sync-hourly | MicroVM Local Sync | 23 * * * * | in 12m | — |
Quorum, membership and the shared key store
Cluster health at a glance: whether there is quorum, which host holds leadership, and how many members are online. The member table shows each node’s database size, Raft term and version, and leadership can be handed to a follower without restarting anything.
- WireGuard mesh SDN, cluster DNS and DHCP on every node
- LeilFS distributed storage across the cluster
- Raft membership with per-node database size
- Force a leader change on a live cluster
| Member | Status | DB size | Raft | Version |
|---|---|---|---|---|
| hoster0103 Leader | Online | 888 KB | term 12 | v9.4.5 |
| hoster0104 | Online | 888 KB | term 12 | v9.4.5 |
| hoster0105 | Online | 888 KB | term 12 | v9.4.5 |
| hoster0106 | Online | 888 KB | term 12 | v9.4.5 |
| hoster-lap-01 | Online | 888 KB | term 12 | v9.4.5 |
Metrics that were already being collected
Every host ships CPU, memory, disk, filesystem, network, systemd, time-sync and SMART metrics, grouped and plotted over any window from thirty minutes to a year. Nothing to wire up: Prometheus and the exporters are part of the install.
- CPU by mode, saturation per core, load average, temperature
- Disk throughput, IOPS, wait time and queue depth
- SMART wear, power-on hours and NVMe media errors
- Power draw per host and per machine, in watts
VMs and microVMs have the same monitoring pages, scoped to the machine instead of the host.
Public names for private machines
Point a domain at a VM or microVM by its short name and the proxy configuration is generated for you. Certificates are automated through your DNS provider, and a domain can apply across every host or be pinned to selected proxy nodes.
- Caddy routes generated from resource names
- TLS automation with stored DNS credentials
- Per-domain host scope
| Domain | State | Routes | Targets | Scope |
|---|---|---|---|---|
| git.example.com | Enabled | 1 route | 1 target | All hosts |
| photos.example.com | Enabled | 1 route | 1 target | All hosts |
| cloud.example.com | Enabled | 1 route | 1 target | All hosts |
| vault.example.com | Enabled | 1 route | 1 target | All hosts |
Who changed what, and when
Every state-changing action, login and denied authorisation attempt, newest first — with the user, the resource path, the response status and the source address. Filter by user, action or group, or show only denied attempts.
- State changes recorded per resource
- Logins and denied authorisation attempts
- Filterable by user, action and group
| Time | User | Action | Resource | Status |
|---|---|---|---|---|
| 03/08 17:55:23 | admin | microvm:manage | Default / hoster0105 / dr-manager-lab | 200 |
| 03/08 17:51:07 | admin | auth.login | /auth/v1/cookie | 200 |
| 03/08 11:04:04 | admin | host:manage | Default / hoster-lap-01 | 200 |
| 03/08 11:01:47 | admin | snapshot:read | Default / hoster0106 | 200 |
| 31/07 13:44:57 | admin | vm:manage | Default / hoster-lap-01 / elite-olive-wolf-001 | 200 |
Start with one server. Add the rest when you need them.
The operating model is a function of how many hosts you run. Nothing has to be decided up front — a single host grows into a cluster without a reinstall.
In every deployment: VM and microVM management, automatic IP provisioning, ZFS snapshots, power tracking, and monitoring with autodiscovery. Each step only adds — nothing is reinstalled.
| What changes | Single server | Multi-server | Cluster |
|---|---|---|---|
| Management interface | CLI | Web UI + API | Web UI + API |
| Replication | Operator-initiated | Operator-initiated | Scheduled, with retention |
| Networking | One host subnet | Independent per-host subnets | WireGuard mesh across hosts |
| Storage | ZFS on the host | ZFS per host | LeilFS distributed, metadata master failover |
| DNS | Per-host resolver settings | Per-host resolver settings | Cluster-wide, updated automatically |
Run the binaries for free. License the source when you need it.
Pre-built binaries are free to self-host at any scale. Production support, a hosted Control Center, source access, and custom development are commercial agreements.
Self-hosted binaries
Grab the pre-built binaries and run Hoster Stack on your own hardware.
- Pre-built binaries — install and run
- Unlimited servers, VMs and tenants
- Full Control Center web UI and CLI
- Snapshots, replication and clustering included
- Power tracking and Prometheus monitoring
Licensed deployment
Run Hoster Stack in production with our team behind it.
- Cloud-hosted Control Center — we run it, you log in
- Or guided self-hosted deployment and configuration
- Full BSL licence to run Hoster Stack on your fleet
- Ongoing platform updates and security patches
- Direct support channel for ops and incident help
Codebase access
Fork Hoster Stack into your own product while we ship features alongside you.
- Everything in Licensed deployment
- Full BSL source access — fork it, brand it, ship it
- Custom feature development by the core team
- Architecture and PR review on your fork
- Direct line to the engineers who wrote it
Business Source License. Commercial agreements cover either a supported deployment or source access with custom engineering. Your infrastructure stays under your control, with support from the team maintaining the platform.
From install media to first VM.
Three steps, each one a command you can copy. Pick a step to watch it run.
Run Hoster on your hardware. Start with one server.
Start a contract
Tell us about your fleet and we will come back to you.