Your whole dev loop, in one private cloud.

Workspaces with AI agents, one-click deploys, hosting and GPUs, built to work together.

AI built in

Claude Code, OpenCode, already in your workspace

Both come installed. Claude Code uses your own Claude account; OpenCode runs on open models on our GPUs.

  • One command

    agent claude or agent opencode, in any terminal.

  • Keeps working when you leave

    Close the tab and it carries on. Pause auto-stop to leave it working overnight.

  • Knows how to ship

    A deploy guide built in, and lab detect drafts your lab.conf.

  • Open models on our GPUs

    OpenCode can use the models running on our RTX 5090s.

One-click deploys

Deploys that look before they leap

Every deploy shows what it will ship and asks first. If your app doesn't come back healthy, the last version returns.

  • See it first

    Preview exactly what a deploy will change.

  • Tested before it ships

    Every commit is tested, private repositories too. One that fails can't be deployed.

  • Automatic rollback

    If the new version doesn't answer, the old one comes straight back.

  • Environments from a branch

    A full copy of your app, with its own address and data, that follows its branch once tests pass.

  • Every part, one click

    Deploy all ships the database, workers and web in order.

Hosting

An address of its own. A copy every night.

Your app gets its own address on the private network. Its data is backed up every night and test-restored every week.

  • Its own address

    /apps/your-app/, live for everyone invited, the moment it's deployed.

  • Backed up nightly

    Copied at 03:30 and checked against the original.

  • Restores that work

    Every Sunday the backups are restored, databases and all.

GPU compute

A GPU on call. The CPU as a fallback.

Send heavy work to an NVIDIA RTX 5090. When the GPU is busy, it runs on the CPU instead.

  • Your code, as a worker

    Your app sends its heavy jobs to a worker of its own.

  • Falls back by itself

    GPU busy? The job finishes on the CPU.

  • One line to ask for it

    TIER=gpu in your manifest.

Private by default

Invited people only. Every app walled off.

Only invited people can reach your apps, so there's no login page to attack. Each app is walled off from the rest.

  • No login page

    The private network is the login. Strangers only ever see this page.

  • Walled off

    Its own CPU and memory limits, and the internet but nothing else.

  • Only what's yours

    You see your apps and your team's, and nobody else's.

Teams

Bring your team. Work side by side.

Bring your GitHub organization. Everyone gets their own workspace, every branch gets its own environment, and you can see who's working on what.

  • From your GitHub organization

    Install Oerba on it, then add members from GitHub.

  • A workspace each

    Everyone codes in their own workspace. Share one to pair.

  • A branch, an environment

    Every branch gets its own live copy, put away once merged.

Sizes

Four sizes. One line to change.

Pick how much of the machine your app can use. Change it any time.

  • Small

    0.5 CPU 512 MB RAM

    A static site or a small API.

    By invitation

  • Medium

    2 CPU 2 GB RAM

    Most web apps, or a JVM, with a database beside it.

    By invitation

  • Large

    4 CPU 6 GB RAM

    Heavy requests, large files.

    By invitation

  • GPU RTX 5090

    4 CPU 6 GB RAM

    Large, plus a worker on the GPU for the heavy work.

    By invitation

Want in?

Oerba is invite-only. Tell us your GitHub account and what you want to build.

Already invited? Switch Tailscale on and open the address in your invitation.