Your own Ethernet, wherever your machines are.
Etorm is a cloud built around the network. Private networks, programs and storage on one API. You get layer-2 segments you can plug anything into, and the services you need sit on those segments, not off to the side.
Start with the network
Most clouds start with virtual machines and add a network around them. We went the other way. A VPC on Etorm is a real Ethernet segment: broadcast works, multicast works, and so do the discovery protocols your gear already speaks.
Machines join it over WireGuard, with VXLAN or Geneve inside the tunnel. A laptop, a server in a closet, a VM at another provider: they all end up on the same segment, with addresses and names handed out by the segment itself.
One API, and you already know it
Everything on Etorm is a resource in a Kubernetes-compatible API. kubectl works. OpenTofu works. curl works.
The web console is just another client of that API, so anything you can click, you can script. Every project gets its own API hostname, its own roles and its own audit log.
Programs that live on the wire
Compute on Etorm is small WebAssembly programs plugged straight into your network, each with its own Ethernet interfaces and addresses. A DHCP server, a firewall, a protocol you made up last week.
They only run when a frame arrives for them. When nothing is happening they can sleep, with their memory saved to storage, and be woken over the network to carry on where they stopped.
Work that survives a crash
Some jobs can't just start over: a provisioning run halfway done, an order waiting three days for someone to approve it. A workflow on Etorm is an ordinary program, compiled to WebAssembly, and every step it takes is written down.
When a process dies or a deploy restarts it, the workflow replays its history and continues from the step it was on.
A graph database with nothing to babysit
Query it with openCypher from your VPC. There is no database server with a disk to size or patch: the data is split into ranges, and every file of every range is an object in storage.
You pick a size class or let it grow with the data. Transactions see a consistent snapshot across all of it.
PostgreSQL without the server
Connect with psql or the PostgreSQL driver you already use. Underneath is our own engine, written in Rust and running as a WebAssembly program on Compute.
A commit is acknowledged once it is written to object storage, so there is no local disk that can take your data with it.
Or just draw it
Some people think in YAML and some think on a whiteboard. The canvas lets you lay out networks and services as blocks and wires.
A controller builds what you drew, and the same picture shows you what is running and what is not. Behind the canvas it is still the same API, so the drawing can be exported and reviewed like any other file.
Also on the platform
Storage
Buckets over plain HTTPS. Every bucket has its own keys, and deleting it makes the data unreadable on the spot.
DNS and DHCP
A DNS zone for every network and DHCP for the ones that need IPv4. Nothing for you to run.
Queues
An MQTT broker on your network, with its state kept in storage.
Load balancing
Load balancers and ingress at L3, L4 and L7, running as processes on your network.
Identity and access
Roles, invitation links, two-step sign-in and an audit log you can actually read.
Architecture
How it fits together, with drawings.
Tell us what you want to connect.
We bring teams on by invitation, and the people who built the platform answer. Say how many machines you have, where they are, and what they need to reach.