You get a kubeconfig. You run kubectl. That's it.
No console to learn, no control-plane fee.
Idle: ~€1.50/month.
Run by Ephais SAS, Paris — company details.
homa is for teams where running your own K8s cluster is too much, and an opinionated PaaS is not enough.
You want kubectl apply without babysitting a control plane, and without a provider charging you for every idle hour.
A stable Kubernetes environment to test, demo and run POCs, without going through corporate ops.
European hosting, run by a French company, with no operational dependency on a US cloud.
You pay for what runs, metered by the minute. No subscription, no tiers. Nothing running → compute at 0 €, only the space fee. Tenant deleted → 0 €, at once.
No tiers. You pay for real usage, and you set a resource cap when you create the tenant — adjustable any time. You will never be billed beyond it.
Monthly ceiling: ~31 € excl. VAT if everything runs 24/7.
Idle, it settles at 1.46 € excl. VAT a month — the tenancy base. Everything else you pay only while it runs.
Google sign-in, then your card — €0 charged today. After that you pay every Monday for what actually ran; idle costs ~€1.50/month. You stop by deleting the tenant.
Create my first tenant →No hidden costs: no bill for network traffic, no bill for API requests, no egress fees. Details in the FAQ.
homa tenant usage <your-tenant> or from your account area.Need a cap above the default limits, or dedicated isolation? Let's talk.
The pipeline that hands you access is documented, scriptable, verifiable.
01 · PROVISIONING
We create a Capsule Tenant bound to your identity, with a namespace, quotas and a RoleBinding. You can't see what isn't yours.
02 · ACCESS
You download your kubeconfig (short-lived token, stable ServiceAccount). Your kubectl talks straight to the Kubernetes API over valid TLS. No tunnel, no proprietary plugin.
03 · OPERATION
Every mutation, every secret read is logged server-side. You can ask for your audit logs at any time.
$ kubectl --kubeconfig=my-tenant.kubeconfig get pods NAME READY STATUS RESTARTS AGE my-app-7d8c9b6f-xz4qr 1/1 Running 0 12m my-worker-5d4f7c8b-jk3lt 1/1 Running 0 12m $ kubectl apply -f deployment.yaml deployment.apps/my-new-service created
Isolated, yes — at the policy level, not the kernel level. Each tenant gets its own namespace, RBAC scoped to it, a ResourceQuota so nobody starves the neighbours, and a network policy that firewalls your egress away from every other tenant's network (private ranges blocked). What you do not get is a private kernel: containers share the host, as on any shared-cluster offering. If your threat model needs hardware isolation, you need a dedicated cluster — and we will tell you so rather than sell you this.
You get a dump of your configmaps + secrets + manifests (kubectl get -o yaml across your whole tenant) and a tarball dump of your PVCs. Then everything is deleted. No hidden retention.
homa is run by one person (Vincent). No contractual SLA at this early stage. Maintenance announced by email 48 h in advance. Support by email, target response within 24 business hours.
Because there you pay for a whole cluster — control plane, nodes, load balancer — even if you use 5 % of it. And you carry the ops work yourself. homa shares one cluster between tenants that don't need a whole one.
Every minute we sample what your tenant really consumes: CPU actually burned, memory in use (working set), plus provisioned storage. We integrate that over time; every Monday we charge the past week × published rates. A pod that is up but idle consumes almost nothing — so you pay almost nothing.
Concrete examples:
Card via Stripe: you register a card (0 € authorisation) when you create the tenant, then an automatic charge every Monday for the past week's usage. Invoiced by Ephais SAS, based in France.
Pick a config, register a card, grab your kubeconfig. All self-service, done in a few minutes — pay for what you use, no commitment.
Create my first tenant