PROVIDERS

A provider is a module, not the unit of your cloud.

Two clouds can look identical and behave nothing alike. One ships DNS, firewalls, private networks and load balancers behind one API; another gives you a server and little else. Flui absorbs that difference in one place instead of scattering it through the CLI, the dashboard and the orchestrator.

Which means a provider does not have to implement everything to be worth integrating — and Flui does not have to pretend every cloud is the same one.

A provider declares what it can do. Flui adapts.

Not the lowest common denominator. Where a provider is missing a capability, Flui says so and falls back — a firewall enforced on the host instead of at the edge, a bootstrap key injected through cloud-init instead of a key registry, DNS left to you instead of pretending it is managed.

Capability	Hetzner	Scaleway	OVHcloud	Your own servers
Machines
Creates servers for you Flui provisions and destroys nodes through the provider API	●	●	●	—
Hourly billing The economic prerequisite for a cluster that buys its own machines	●	●	●	—
Resize a node in place Change a machine to a larger shape without rebuilding the cluster	●	●	—	—
Stop a node without destroying it	●	●	—	—
Live stock, per region Whether the shape you picked is actually available before Flui asks for it	●	●	—	—
Network
Private networking Nodes talk over a private network instead of public IPs	●	●	●	◐
Managed DNS zones Flui creates and reconciles the records your applications need	●	●	—	—
Firewall at the provider edge Rules enforced in front of the machine rather than only on it	●	●	—	—
Storage
Block volumes for the cluster Created, attached and grown by Flui — this is what shared cluster storage sits on	●	●	●	—
Object storage Flui can create A bucket and keys for backups, made for you rather than by you	◐	●	●	—
Account
SSH key registry Keys held on the account and handed to a machine at creation	●	●	●	—
Inference from the same account Models Flui can reach with the credential it already holds for compute	—	●	—	—

● supported · ◐ partial · — not supported today
Hetzner's object storage wants a second set of credentials; Scaleway's reuses the compute one, and OVHcloud mints an S3 key from it. OVHcloud's two gaps are gaps in the cloud, not in Flui: its security-group quota is zero in practice, so the rules land on the host exactly as they do on a machine of your own, and DNS stays with whoever holds your zone. On your own machines the private network is whatever you wired.
Two things are deliberately absent. Provider snapshots and provider load balancers are things these clouds sell, and things Flui does not drive on any of them — backups are Flui's own, taken from inside the cluster to any S3 endpoint, and ingress is Traefik running on your nodes.

Partial support is a first-class outcome.

A provider need not do everything

OVHcloud buys and destroys machines on command, bills them by the hour, and gives the nodes a private network to talk on — and exposes no firewall Flui can drive and no DNS it can reconcile. That is not a broken integration. It is a provider participating at the level it supports, on the rows where it can.

Your own machines are a provider too

A VPS, a rack, a machine under a desk. Bring-your-own-server has no provisioning API by definition, so nodes and firewalling are driven over SSH instead. Everything above — applications, databases, scaling, observability — is the same product.

The account stays yours

Flui holds an API token you issued and can revoke, encrypted at rest. Servers are created on your account and billed to you. Nothing is resold through us, and there is no Flui-owned layer between you and the invoice.

Where Flui runs

Where Flui runs today.

Same flui.yaml, same operating model, whether the machines come from an integrated provider, a VPS, bare metal or your own Ubuntu servers. Every pin is a region Flui buys machines in today — and anywhere else, it installs on machines you bring.

Hetzner Scaleway OVHcloud

Need a different provider or BYOS path? Get in touch →

Not yet

What "composable" does not mean today.

A single application cannot yet draw compute from one provider and storage from another. Composition today happens between clusters, not inside one.
Only Hetzner, Scaleway and OVHcloud provision nodes. Everywhere else, you bring the machines.
On OVHcloud, Flui writes no DNS and no edge firewall rule. Point a zone held on Hetzner or Scaleway at the cluster, or manage the records yourself; the firewall is enforced on the machines instead.
Nodes run Ubuntu 24.04 on x86_64. ARM is not supported.
There is no hyperscaler module. It is architecturally ordinary — a provider is a module against a fixed interface — but it is not written.

The direction is on the vision page, kept separate from this one on purpose.

For partners

Build your cloud offering on Flui.

If you operate cloud infrastructure but lack a modern software layer, Flui is the platform your customers expect. Open source, customizable, and engineered for production.

White-label friendly

Brand the platform as your own. Flui is the engine, you're the offering.

Engineered for production

Tested, documented, and built on K3s, Postgres and the same battle-tested open-source stack you already trust to run real workloads.

Custom integrations

Need a specific provider, custom workflow, or tailored deployment? Let's design it together.

Talk about partnering

The capability interface, and what it takes to add a provider, are documented. Read the provider layer →