Three products, one contract.
Zephyr owns the commercial facts, kstack owns what is placed on a machine, Standby owns the physical facts about the machine. Each is bought on its own and runs on hardware you already have. The reason they work together is not an integration — it is that none of them keeps a half-right copy of another's.
What each one costsZephyr
Billing, provisioning, domains and support for a hosting company. This shop is running on it.
- A paid invoice becomes a working account and a missed one suspends it, with no manual step in between and nobody retyping an order into a control panel.
- Double-entry underneath proration, credit notes, partial payments, dunning and renewals over a real ledger, exported as CSV, XLSX or QuickBooks IIF.
- Tax handled rather than approximated multi-currency with stored FX rates, VAT zones, live VIES validation and reverse charge.
- Provisions into what you already run kstack, cPanel, Plesk, DirectAdmin, Proxmox, Hetzner Cloud, OpenStack, and dedicated hardware over Redfish.
- Your servers, your PostgreSQL, your data and no per-seat charge. If the licence lapses the software keeps running; only updates stop.
- The public site included pages, blog, changelog and help centre, and a shop priced from the same catalogue billing reads.
kstack
The control panel your customers log into, and the agent that keeps each machine matching what you asked for.
- The control plane never touches your servers it writes desired state and returns. Each node pulls its manifest, compares it against what is actually there, applies the difference and reports back.
- Drift is corrected, not discovered a config edited by hand at two in the morning is put back on the next cycle, and you hear it from the report rather than a customer.
- Thirteen resource types, one lifecycle websites, DNS zones and records, mailboxes and mail domains, databases and users, certificates, FTP accounts, cron jobs and backup policies.
- The website is the isolation unit one unprivileged, id-mapped container per site, so a customer with five sites has five boundaries rather than one.
- Nodes keep serving without the control plane they hold their manifest, so the sites on them do not depend on the panel being up.
- The panel is a client of the public API no private endpoints it uses and you cannot. A design rule rather than a roadmap item.
Standby
On-call, monitoring and security operations across every machine you run, whoever put them there.
- A status page on its own deploy separate infrastructure from the estate it reports on, serving a static snapshot so an audience arriving at once reads a file rather than a render.
- CVEs matched to what is actually installed each host's real package inventory at the version it actually has. A stale inventory is refused rather than matched.
- Known-exploited first KEV-flagged findings drive a detect-to-fix loop with the package upgrade that closes them, so triage starts with what is being exploited.
- Escalation that stops when it is handled chains by severity and label, on-call schedules, overrides and silences. A fleet-wide event rolls up and pages once rather than two hundred times.
- Backups reported as proven or not proven restore drills replay the snapshots on a schedule and record the count, so “when did you last test that” has a date.
- One agent across the fleet Debian, Ubuntu, RHEL and SUSE, a musl build under OpenRC for Alpine, and Windows. Where a platform cannot do something it says so.
Side by side
| Zephyr | kstack | Standby | |
|---|---|---|---|
| What it owns | The commercial facts: catalogue, invoices, the ledger, the customer. | What is placed on a machine and its lifecycle: sites, zones, mailboxes, databases, certificates. | The physical facts about a machine: reachability, patch state, CVE exposure, backup proof. |
| Who buys it | Anyone selling hosting, domains or subscriptions. | Anyone running the machines their customers sit on. | Anyone answerable for a fleet at three in the morning. |
| From | £89 a month | £39 a month | £149 a month |
| What the first tier covers | One install. No per-seat charge. | Up to ten nodes, with customers and sites unlimited at every tier. | Up to twenty-five endpoints, with the status page included rather than sold separately. |
| What it runs on | Your servers and your own PostgreSQL. | Debian stable nodes with cgroup v2 that you already run. | One agent on Debian, Ubuntu, RHEL, SUSE, Alpine and Windows. |
| Whether it needs the others | No. It provisions into cPanel, Plesk, DirectAdmin, Proxmox, Hetzner and OpenStack as readily as into kstack. | It implements no billing of its own, deliberately — so it needs something, and Zephyr needs no integration work. | No. It watches machines whoever put them there. |
Why they are separate
Bought one at a time, on purpose.
This is not a suite with a bundle price. There is no lock-step release train and no discount for taking all three — what they share is an agreement about who owns which fact, and that agreement is just as useful when you only own one of them.
One owner per fact
Zephyr provisions into kstack: a paid invoice becomes an account, a subscription and its resources against the plan bought. kstack raises alerts as incidents and pushes its serving probes; Standby returns what the machine is actually doing, which kstack displays rather than measuring a second time.
The catalogue is one thing
kstack publishes plan kinds, sizes and measured facts per location. Zephyr sells that, rather than a list somebody retyped — which is why a location only appears for sale when a node there can actually serve it.
Updates arrive signed
All three come down the same release edge: licence-gated, Ed25519-signed and verified against a pinned key before anything is swapped in. The install command is personalised per licence and appears in your portal after purchase.
The panel is a client of the API
kstack uses no private endpoints you cannot. If the panel can do it, so can your script. That is a design rule rather than a roadmap item, and it is what makes replacing any one of these three possible later.
Rehearse before you commit
Zephyr's importer reads a WHMCS or HostBill database directly, read-only and dry-run by default, so you check the totals against your own clients before anything is written. Standby is best started on a handful of hosts for a week.
We are the first customer
lbreeze's hosting runs on all three. When something is wrong with one of them we find out the way you would, and the fix is a release rather than a ticket raised with somebody else.
Frequently asked questions
Do I have to buy all three?
No, and most people do not start that way. Each is licensed on its own. kstack with somebody else's billing works; Zephyr with somebody else's control panel works, because provisioning goes through a public, documented API rather than a private handshake.
Do I have to use Zephyr for billing if I run kstack?
You need something, because kstack implements no billing at all — that is deliberate rather than unfinished. Zephyr is built to its provisioning contract and needs no integration work. Anything else means writing against the provisioning API yourself, which is public and documented.
What happens if a licence lapses?
For Zephyr this is stated plainly: the software keeps running and only updates stop. If you need the same commitment in writing for kstack or Standby, ask before you buy rather than after.
How do updates reach my machines?
Through the same signed release edge for all three — licence-gated, Ed25519-signed and verified against a pinned key before anything is swapped in. Nothing is fetched from a mirror and trusted because it downloaded cleanly.
Where is the install command?
In your portal after purchase. It is personalised per licence rather than a public one-liner, which is also how the licence gets onto the machine.
Can I see it against my own data first?
For Zephyr, yes, and it is the most useful first conversation: a dry-run import of your existing system so you are looking at your own clients and invoices rather than a demo. For Standby, put the agent on the hosts you would miss and let it run for a week.
What if I am running more than the first tier covers?
Every product is priced in flat tiers rather than per unit, and the tiers are on each product's pricing tab. Beyond the top tier of any of them, it is a conversation rather than a published number.
Not ready to run a platform
A reseller plan is the version of this with the machines, the patching and the monitoring already done. Your clients, our infrastructure, from £39 a month.
Tell us what you are running now.
The panel, the billing system and roughly how many customers and machines. We will tell you which of the three is worth starting with and which one you should leave alone for the moment.
Book a walkthrough