Runners
The machines your columns work on: a cloud machine ORVEN runs for you, or your own laptop, server or VM.
What a runner is
A runner is a machine running orven-harness. Agents, terminals, processes and databases run there, in your files, with your tools. The board is where you see them and work them, from any browser.
There are two kinds, and one board can use both: cloud machines ORVEN runs for you, and machines of your own. Both are listed in Settings, Runners, and their columns work the same.
Cloud machines
Your plan comes with a cloud machine. It's set up when you sign up, and a few minutes later it joins your runners as cloud.
- Claude Code and Codex are installed, ready to sign in from Settings, Runners.
- You and your agents have root on it: install anything, run Docker.
- Its disk is backed up every day.
- It runs in Germany or Finland.
The ORVEN plan. Every column, a small cloud machine, and your own machines, as many as you like.
Get ORVEN| Machine | Cores | Memory | Disk | Price |
|---|---|---|---|---|
| SmallWith your plan2 cores · 4 GB memory · 40 GB disk | 2 | 4 GB | 40 GB | Included |
| SmallAdded to your plan2 cores · 4 GB memory · 40 GB disk | 2 | 4 GB | 40 GB | $15/month |
| MediumAdded to your plan4 cores · 8 GB memory · 80 GB disk | 4 | 8 GB | 80 GB | $25/month |
| LargeAdded to your plan8 cores · 16 GB memory · 160 GB disk | 8 | 16 GB | 160 GB | $45/month |
| Your ownLaptops, servers, VMs | Any size, as many as you like | Included | ||
More cloud machines
Add a machine, in Settings, Plan, buys another. It's billed beside your plan from the day you add it, and joins your runners as cloud-2, cloud-3 and so on. Cancel one any time in billing.
An added machine can grow to a bigger size from its row in Plan. It's off for a few minutes while it changes, and the new price starts then, prorated. A machine can't shrink, since its disk can't.
What's new to a cloud machine's setup, a tool or a fix, reaches it by itself, between your agents' turns.
Rebuild, in Plan, starts a machine's system over and keeps your work. Everything in your home folder stays, your sign-ins with it, and so do Docker's images and volumes; what you installed or changed outside them goes. It's off for about ten minutes, and its columns carry on where they were.
Recycle, in Plan, swaps a machine for a fresh one. The old one is deleted, disk and backups and all, and its columns start over on your default runner.
If your plan ends, its machines are switched off and kept for a week, in case you come back. Then they're deleted.
Your own machines
Any machine you can run a command on can be a runner: your laptop, a server, a VM. macOS or Linux, on Intel, AMD or ARM. They come with your plan, as many as you like.
Behind a router or a firewall is fine. The runner connects out to ORVEN, so nothing needs to reach it.
Connecting one
- In Settings, Runners, press Add runner.
- Run the command it shows on the machine. It installs orven-harness in
~/.orvenand makes the machine one of your runners. The code in it works once, for ten minutes. - The machine shows up in Runners as soon as it connects, with the coding agents installed on it.
orven-harness runs as a service of your user, so the runner is back after a restart. Remove runner, in its row, takes it off your runners.
Your first runner is the default: a new board's first column starts there, in its working directory, and the columns after it start where you were working. Make default picks another. A runner you can reach directly can also be connected by address, with its name, address, token and working directory.
Coding agents
A runner lists the coding agents installed on it: Claude Code, Codex CLI, Cursor CLI, Grok Build, Kimi Code, OpenCode, Amp and Pi, each with its version and whether it's signed in. One that needs signing in has Sign in, which opens its sign-in page and finishes with the code it gives you. No terminal needed. The runner keeps them up to date.
Working across runners
Your runners are one network, with nothing to set up. An agent on one works on the others as if it were there: orven on runs a command on another runner, orven fs and orven cp read and copy files, and orven forward reaches a port, without ssh. Runners link directly when they can, and through ORVEN when they can't.
Picking another runner on an agent's runner chip moves it there once its turn ends, with its conversation. Its folder stays where it was: the agent is told what it had, the branch, commits not pushed and uncommitted changes, and brings what it needs.
Mark a runner Protected in Settings, Runners, like a production server. Agents on other runners read it freely, and ask you before anything that would change it; it can't reach other runners itself. Between runners, in the same place, lists what each runner did on another.
Offline and updates
Columns on a runner that's offline say so. ORVEN keeps trying to reconnect, and a draft you were typing stays where it is.
Runners update themselves without ending anyone's work: turns, shells and processes that are running finish on the release that started them, and new work goes to the new one.