Getting started

From checkout to your first agent session.

You are buying one Linux workstation in the cloud, yours alone, with the six agent CLIs already installed. Here is exactly what happens between paying and typing claude into a terminal.

What you actually get

One machine. Not a container on a shared box, not a sandbox that resets, not a web IDE with a terminal bolted on. Your plan buys a dedicated virtual machine with its own persistent disk, running a full Linux userland where you have sudo.

You reach it through a browser at your own hostname. Inside that browser window is the HolyCode workspace app: a file explorer, an editor, and as many terminal sessions as you want. The terminals are real PTYs on the real machine: there is no simulation layer between you and the box.

An "agent" here is just a CLI you run in one of those terminals. Your terminal session list is your agent list. See Connecting your AI agents.

The five steps

  1. You pick a plan Pick a tier. Monthly, no annual option, and no general free trial: the product is paid from day one unless you were given a trial code. We sell worldwide, in US dollars. You agree to the Terms of Service by creating an account and subscribing.
  2. You choose two things, before you pay We ask for the region your machine runs in and the address it answers on, something like yourname.app.holycode.cloud. Six regions: Chicago, Ashburn, San Jose, London, Frankfurt and Sydney. If you don't mind, one click takes the defaults (Chicago, and an address we generate) and you're straight through to checkout. Both choices are permanent, so we ask before your money moves rather than after: your disk is created in the region you pick and can't be moved, and the address becomes your machine's DNS record, its TLS certificate and its sign-in identity.
  3. Your machine builds itself Checkout hands off to our own systems, which create your virtual machine in the region you chose, size it to your tier, attach the volume, point DNS at the address you chose and issue its TLS certificate (no human in the loop). It's usually live within minutes. The promise we make is the floor, not the typical case: delivery within 24 hours of your payment settling. Because we take both details before checkout, there is nothing we need to come back and ask you for. If anything in that chain fails, a person finishes the job by hand rather than leaving you waiting.
  4. You get the welcome email It contains your workstation URL and a note on how to authenticate each agent CLI. There is no password to send: you sign in at app.holycode.cloud with the account you already made, and open the workstation from your dashboard.
  5. You open it and start a session Sign in, click Start a session, and you have a shell. Type the name of the agent you want. First run walks you through that CLI's own sign-in.
If we miss the 24 hours, you get a full refund. That's not goodwill, it's in the Refund Policy. If we're running late you'll hear from us before the deadline rather than after it, and you can take the refund or keep waiting. Your call.

Your first minute inside

you@holycode: session 1
# the six agents are already on the box
you@holycode:~$ claude
# first run signs you in with your own Claude plan or API key

you@holycode:~$ cd ~/workspace && mkdir my-project
# projects are just folders. make as many as you like.

There is nothing to install and no project limit. A "project" is a directory on your volume. Create them with mkdir. Disk is one shared quota across everything you keep, so a big monorepo and forty small repos both just spend from the same number.

Sleeping and waking

Your workstation suspends when it's idle and wakes when you come back. A warm resume takes about a second, with your processes and memory intact: the session you left running is the session you return to.

The limit, plainly. After a long idle stretch the platform sometimes falls back from suspend to a full stop. That wake is a cold boot (think tens of seconds, not one second), and running processes don't survive it, though everything on disk does. It's uncommon, and it's the same trade every suspend-based platform makes. What survives what is spelled out in Security basics.

What isn't automated yet

Being straight about the beta:

Next

Get your agents signed in: that's the part that makes the machine worth having.

← All docs Connecting your AI agents →