Create your fleet.
Sign up and choose a name. Your fleet gets its own hosted console, ready for its first machine.
YOUR CONTROL PLANE
acme.odalr.com ↗ OPEN BETA FREE TO EXPLORE ↗
YOUR INFRASTRUCTURE / REIMAGINED
Your hardware.
One cloud.
A cloud on machines you own,
run for you.
Bring the server in your rack, the workstation in your office, the box under your desk. Deploy across them all. We take care of the coordination.
No credit card. Start with one Linux machine.
ACME / FLEET OVERVIEW CONNECTED
MANAGED BY ODALR
One control plane Placement · Networking · HTTPS
01 / HOME
home-01
8 cores 32 GB
◇ web
02 / COLOCATION
rack-01
32 cores 128 GB
◇ postgres
03 / OFFICE
studio-01
16 cores 64 GB
◇ llama3.2
↑ YOUR MACHINES. YOUR WORKLOADS. WIREGUARD MESH
DEPLOY TO THE FLEET
$ odalr run web --image nginx:alpine
✓ web started placed on home-01
03 MACHINES 56 CORES 224 GB MEMORY
EXAMPLE FLEET
ANYWHERE YOUR HARDWARE LIVES
Home lab + Office + Colocation + Existing cloud → One fleet
01 / THE CLOUD WORKFLOW WITHOUT THE WEEKEND OF SETUP
Set up once.
Deploy whenever.
You already have the compute. odalr is the layer that makes it useful together. One agent per machine. One place to run everything.
Bring your machines.
Run the join command from the console on each Linux machine. They connect through NAT over an encrypted private network.
ONE COMMAND PER MACHINE
✓ rack-01 joined the fleet
Give them something to do.
Use the console or CLI to deploy. odalr picks a machine, prepares the runtime and keeps the workload running.
YOU SAY WHAT. WE FIND WHERE.
$ odalr run web --image nginx
LEAVE THE NETWORKING TO US Public IPs Port forwarding Manual VPN setup Just connect. ↗
02 / A WHOLE CLOUD ON YOUR HARDWARE ONE CONSOLE. ONE CLI.
Small commands.
Real possibilities.
The app, its database, the model behind it. Give your whole stack a home on infrastructure you control.
COMPUTE
From image to internet.
Ship an image. odalr finds a machine with room, keeps it running, and gives it an HTTPS address when you ask.
$ odalr run web --image ghcr.io/acme/web$ odalr expose web:8080✓ https://web.acme.run.odalr.com DATA
Your database. Your disks.
Postgres, MariaDB, Valkey or MongoDB, with persistent storage, a generated password and nightly snapshots. Add replication when you need it.
$ odalr db create main --engine postgres --storage 20G✓ main created · nightly backups scheduled INFERENCE
Give that GPU a day job.
Serve a model on your own hardware through an OpenAI-compatible API. odalr selects a suitable runtime and keeps the weights on a persistent volume.
$ odalr model run llama3.2✓ llama3.2 deployed · OpenAI-compatible endpoint VIRTUALIZATION
A whole OS. Right at home.
Run KVM-backed Linux virtual machines on Intel, AMD or ARM. Reach them over the same private network, with your existing SSH keys.
$ odalr vm build-box --memory 8G --disk 100G$ odalr ssh build-box -l ubuntu✓ build-box ready · private overlay connected ORCHESTRATION
Real Kubernetes. Less upkeep.
Create an upstream k3s cluster on VMs placed across your fleet. Use kubectl and Helm, with integrated volumes, identity and public routes.
$ odalr cluster create prod --ha✓ prod created · upstream k3s SERVERLESS
Just bring your handler.
Deploy a function or a whole Web fetch app. Use Lambda-style handlers, Hono, Elysia or Python ASGI on your own compute.
$ odalr fn deploy resize --dir ./functions/resize --handler index.handler✓ resize deployed AND THE REST OF YOUR STACK
S3 object storage Volumes & snapshots Private networks Container registry Metrics & logs Secrets & identity Jobs & autoscaling
03 / A CLEAR DIVISION OF LABOUR MANAGED WHERE IT HELPS. YOURS WHERE IT MATTERS.
Your cloud.
Actually yours.
Your applications run on your machines. Your databases and files live on your disks. We run the control plane that brings it all together.
Get into the details ↓ WE LOOK AFTER
The coordination.
Console & API · Scheduling · Private networking
Certificates · Signed agent updates
ONE CONNECTED SYSTEM ↓ ↑
YOU OWN
Everything you’re building.
Applications · Databases · Models · Files
Disks · Container images · The actual machines
04 / FAMILIAR BY DESIGN BRING YOUR STACK WITH YOU
A new home.
For the tools you know.
Keep the images, manifests, clients and workflows your team already uses. Build on established protocols and upstream software.
Ship
Deploy what you already build.
- Docker & Compose
- Kubernetes & Helm
- Lambda handlers
Connect
Use familiar clients and drivers.
- S3
- Postgres · MySQL · Redis · MongoDB
- OpenAI API
Operate
Keep your everyday tools close.
- SSH · scp · rsync
- Prometheus & Grafana
- OIDC single sign-on
BUILT FOR YOUR KIND OF HARDWARE
Start where you are.
A team with a rack
Turn the servers in the office into staging and production, with the same deploy flow you would have on a cloud.
A lab with GPUs
Serve a model from the workstation under the desk at an OpenAI-compatible URL the whole team can use.
An agency
Host client sites on hardware you control, each with its own domain and certificate, all from one console.
A home lab
Run your own services at home and reach them from anywhere, without opening a single port on your router.
OPEN BETA / YOU’RE INVITED
Your hardware is ready.
Your hardware is ready.
Your cloud can be, too.
Create a fleet, connect a machine, deploy something.
Create your fleet ↗ FREE DURING BETA
$0/ while in beta
No card. No usage limits during beta. We’ll announce pricing well before anything is charged.
Start with one machine. Grow from there. 05 / THE DETAILS
Good questions.
Good questions.
Straight answers.
What machines can I add?
Any Linux machine you can run a command on: a server, a workstation, a mini PC, or a VM at another provider. Intel, AMD and ARM all work. It does not need a public IP or any network setup.
Do I need to open ports or set up a VPN?
No. Each machine connects out to odalr, and your machines reach each other over an encrypted private network that works through any router. Visitors to your public sites reach them without you opening anything.
Where does my data live?
On your machines. odalr keeps the description of your fleet: what should run, where, and with which settings. Your applications, disks, databases and files are on your hardware only.
What happens if odalr is unavailable?
Your machines keep running whatever they were running. You cannot make changes until we are back, and traffic to public addresses passes through odalr today, so those would be unreachable in the meantime.
Can I host the control plane myself?
The control plane is hosted and managed by odalr; self-hosting it is not offered today. Your machines run the agent and your workloads. We maintain the console, API and coordination.
Can I use my own domain?
Yes. Add one DNS record to prove it is yours, and odalr issues and renews the certificates for it.
How do I get the odalr command?
One line: curl -fsSL https://odalr.com/cli.sh | sh. macOS and Linux, Intel and ARM, and it does not need root or a package manager.