A remote computer for Claude
Your own Linux machine with Claude Code on it, working around the clock. Sign in to your Anthropic account and start.
How it works
One button in the dashboard. We build the machine and write to you when it is ready.
What arrives
A machine of your own
Ubuntu LTS, at least 1 vCPU, 4 GB of memory, 40 GB SSD and an IPv4 address of its own. Root is yours.
What this means
A virtual machine of your own, not an account on somebody else's. Typically Linux (Ubuntu LTS today), with at least 1 vCPU, 4 GB of memory and 40 GB of SSD, usually in a European datacentre, and a public address of its own, an IPv4 at the moment.
You are the only resident and you have full root access, so the machine can be configured however your work needs. There are two ways in: the terminal built into your dashboard, which needs nothing installed on your side, or ordinary SSH once you have added your key.
Claude Code, installed and updated daily
The way Anthropic publishes it, updating itself every day. A working folder waits with a starter CLAUDE.md.
What this means
Claude Code is already installed when the machine reaches you, the way Anthropic publishes it, and it updates itself daily. A working folder is waiting with a starter CLAUDE.md describing the machine, which is usually the first file people edit.
Around it we install the tools an agent needs for office work rather than only the ones a programmer needs: Node and Python, a headless Chrome, spreadsheet and document libraries, PDF tools and text recognition in English and Russian. The exact list moves as we learn what the cases actually call for.
Your own Anthropic account
Claude runs on your account and Anthropic bills you directly. We never see the key and add no markup.
What this means
Claude runs on your Anthropic account, never on ours. You sign in once from the browser terminal and pay Anthropic directly for whatever your agents use: no markup from us, no shared pool, no quota that we set.
We never see, hold or store those credentials, which is also why nobody here can spend on your behalf or read what you say to the model. We recommend a Claude Pro or Max plan rather than API billing: that is what Claude Code is priced for, and an agent working through the night costs far more per token on the API.
Two ways in
A terminal inside your dashboard, with nothing to install. Ordinary SSH once you add your key.
What this means
A virtual machine of your own, not an account on somebody else's. Typically Linux (Ubuntu LTS today), with at least 1 vCPU, 4 GB of memory and 40 GB of SSD, usually in a European datacentre, and a public address of its own, an IPv4 at the moment.
You are the only resident and you have full root access, so the machine can be configured however your work needs. There are two ways in: the terminal built into your dashboard, which needs nothing installed on your side, or ordinary SSH once you have added your key.
Office tools, not only code
Node and Python, a headless Chrome, spreadsheet and document libraries, PDF tools, text recognition in English and Russian.
What this means
Claude Code is already installed when the machine reaches you, the way Anthropic publishes it, and it updates itself daily. A working folder is waiting with a starter CLAUDE.md describing the machine, which is usually the first file people edit.
Around it we install the tools an agent needs for office work rather than only the ones a programmer needs: Node and Python, a headless Chrome, spreadsheet and document libraries, PDF tools and text recognition in English and Russian. The exact list moves as we learn what the cases actually call for.
A mailbox and a bot of its own
Forward a document to the machine's address and the agent starts on it. Telegram waits for a bot token.
What this means
The machine is set up for work that happens while nobody is watching. It gets a mailbox of its own, so forwarding an invoice or a report to it is how a job starts. Only senders on the list you control can give it instructions; anything else is treated as material to work on and never as an order.
Telegram is installed and waiting for a bot token, which you create yourself and paste in as one line. Chrome is on the machine and its screen is visible from your dashboard, so an agent can browse, fill in forms and take screenshots while you watch it happen.
Who writes the agents depends on the plan. On Start and Basic what they do is yours to write, and on Basic we also watch that they break nothing; on PRO we set them up and keep them current. Either way the machine comes with a run cap and a timeout, so a loop between two automated systems cannot spend a month of your Anthropic budget overnight.
Plans
Start
A remote computer with Claude Code on it, and one masterclass a month
€19 per month
€193.80 per year
€16.15 a month, billed once a year, 15% off
- Remote computer
- Pre-installed Claude Code
- Your own Anthropic account
- Agents you set up yourself: email, Telegram, browser
- Masterclasses: 1 a month, your pick
Basic
The same computer, the whole masterclass library, and backups that run themselves
€39 per month
€397.80 per year
€33.15 a month, billed once a year, 15% off
- Remote computer
- Pre-installed Claude Code
- Your own Anthropic account
- Agents you set up, watched by us so nothing breaks
- Masterclasses: Full
- Backups: Daily
PRO
Everything in Basic, plus the recordings live and half an hour with the team every quarter
€79 per month
€805.80 per year
€67.15 a month, billed once a year, 15% off
- Remote computer
- Pre-installed Claude Code
- Your own Anthropic account
- Agents we set up and keep current
- Masterclasses: Full, with live participation
- Backups: Daily
Before you order
Is a Claude subscription included?
No. Claude Code signs in to your own Anthropic account and Anthropic bills you for what your agents use; we add nothing on top. We recommend a Claude Pro or Max plan rather than API billing: that is what Claude Code is priced for, and an agent working through the night costs far more per token on the API.
Can I use my existing Anthropic account?
Yes. Claude Code on the machine signs in to whichever Anthropic account you already have, on the same Pro or Max plan you use on your laptop; nothing new has to be opened. You sign in once from the browser terminal, and the credential goes from Anthropic to your browser to your machine without passing through us.
How soon is the machine ready?
Paying sets nothing up. When you press the button on your dashboard asking for it, we build it and mail you, usually within a few hours. Until you press it, the statutory 14 day withdrawal period is untouched (clause 10 of the Terms of Service).
Do I need to know Linux or SSH?
No. The terminal is inside your dashboard, in the browser, with nothing to install: open it, type claude and follow the sign-in link. SSH is there for whoever wants it, once you add a key on the server page, and it is never required.
Can anyone at combobulating read what is on it?
Not in the ordinary course of business, and nothing of ours monitors it. We hold root because we run the machine, and a person looks only in four cases: you ask us to, you share a project with our team on a plan that includes it, we have a specific reason to suspect a security incident, or a UK or EU court or authority requires it. Clause 8 of the Terms of Service names each one and its limits.
What may it not be used for?
No information security work of any kind, no spam, no reselling the machine as hosting, and nothing unlawful, including anything that breaks UK, US or EU sanctions rules. The Acceptable Use Policy is the complete list; if you are in any doubt, ask us before you order.
Are there backups?
On Basic and PRO the whole machine is snapshotted once a day and about a week of snapshots is kept. It is a whole-disk image: restoring puts the machine back as it was that night. Start has no backups, and moving up starts them from the next nightly run.
What if an agent loops all night?
The machine comes with a run cap and a timeout, so a loop between two automated systems cannot spend a month of your Anthropic budget overnight. Only senders on a list you control can give the agent instructions; anything else is material to work on, never an order.
Can I run other agents on it?
Technically yes: it is a Linux machine and it is yours. We cannot support anything else properly yet, so whatever you install beside Claude Code you run on your own.
How do I pay, and can I cancel?
In euros, by card, through Stripe; any international Visa or Mastercard works and your bank converts. Cancel whenever you like: the machine keeps running to the end of the period you paid for, and on that day it and its backups are deleted, so copy off what you need first.