A machine for Claude Code that never closes its lid
$
Give your coding agent somewhere permanent to work. Long refactors keep running after you disconnect, your API keys live on the server instead of six laptops, and every session leaves a record.
A large refactor can take an hour of tool calls. On a laptop that hour is hostage to sleep settings, VPN drops and battery. Here the session runs under a process supervisor on the server: you attach to watch, detach when you want, and the agent carries on. If your connection dies mid-run, nothing is lost and you reattach to the same session.
02
Credentials stop travelling
The usual pattern is an Anthropic key in a dotfile on every developer's laptop, plus a Git token beside it. That is several copies of a credential on several machines that leave the building. On a hosted machine the keys sit in a secret store, get injected into the session environment at run time, and are rotated centrally. Revoking someone's access is one change, not a request that they delete a file.
03
You can answer 'what did it change?'
Agents write code. Sooner or later someone asks which commits came from an agent, what it was asked, and what it touched. We keep the prompt, the tool calls, the file diffs and the exit status for every session, so that question has an answer rather than an argument. On Scale the history is kept for a year.
04
Agents can run on a schedule
Once the machine is always on, some jobs stop needing a human to start them. Run an agent at 2am to triage the overnight test failures and open a draft pull request. Trigger one from a webhook when a dependency alert lands. This is the part people do not anticipate when they first move off a laptop, and it is usually what they end up valuing.
the options
Where should the agent actually run?
Claude Code runs anywhere there is a terminal. The question is what happens to a forty-minute job when you shut the laptop and get on a train.
where-should-it-run.diff
Long-running jobs
On your laptop: Stop when the machine sleeps
A bare cloud VM: Keep running, if you set up tmux yourself
The Email Shop: Keep running — sessions are supervised and resumable
Where your API key lives
On your laptop: On every laptop that runs it
A bare cloud VM: In a file you put there
The Email Shop: In a managed secret store, injected at run time
Setup
On your laptop: Per developer, repeatedly
A bare cloud VM: You build the image and keep it current
The Email Shop: Node, Python, Go, Docker and the CLI, already there
Who else can pick it up
On your laptop: Nobody — it is your machine
A bare cloud VM: Anyone with the SSH key
The Email Shop: Named team members, each with their own login
Record of what changed
On your laptop: Your shell history, until it rotates
A bare cloud VM: None by default
The Email Shop: Every session logged with commands, diffs and timings
Cost of an idle machine
On your laptop: Nothing
A bare cloud VM: Full price, running or not
The Email Shop: From £12.99/month flat
If you only use Claude Code in short bursts while you are sat at your desk, your laptop is the right answer and you do not need this.
Built for agents that work unattended
The difference between a general-purpose VPS and this is everything below, already configured.
Resumable sessions
Start a job, disconnect, come back tomorrow. Sessions run under a supervisor, so a dropped connection does not kill the work or lose the output.
Keys the agent can use but you cannot leak
Anthropic keys, Git tokens and deploy credentials live in a secret store and are injected into the session. Nothing is written into a dotfile on a laptop.
Repositories already cloned
We set up deploy keys and clone the repositories your team works on, so a new session starts inside the project instead of starting with git config.
A record of every session
Prompts, tool calls, file diffs and exit codes, kept and searchable. Useful for review, and necessary the first time someone asks what the agent changed.
Toolchains preinstalled
Node, Python, Go, Rust, Docker and the usual build tooling, kept current. The agent can install more, and we snapshot before it does.
Scheduled and triggered runs
Run an agent nightly against your test suite, or on a webhook when CI fails. The same machine, doing work while nobody is watching.
Prices exclude VAT. No setup fee, no minimum term.
Frequently asked questions
Yes. You bring your own Anthropic account and key, and your model usage is billed by Anthropic directly to you. We host and operate the machine the agent runs on; we do not resell model access or add a margin to your token spend.
No. Claude and Claude Code are Anthropic's products. We are an independent UK hosting provider offering managed infrastructure to run the Claude Code CLI on, in the same way we host any other software. We are not affiliated with Anthropic.
Yes, using deploy keys or a machine user that you control and can revoke. We will set this up with you during onboarding. You decide which repositories the machine can see, and you can change or withdraw that access at any time without involving us.
The machine is yours and isolated from other customers, so the blast radius is your own workspace. We snapshot nightly and before any package installation, so a bad run can be rolled back. Beyond that it is the same as any developer environment: give it credentials scoped to what it should be allowed to touch, and review what it produces before it merges.
Solo is intended for one developer. Team supports up to five named logins and three concurrent sessions, and Scale has no login limit with ten concurrent sessions. Each person signs in as themselves, so session logs show who ran what.
Yes. It is a normal Linux server with root access, so anything you can install, you can run. Plenty of customers run Claude Code alongside a scheduler, a scraper or their own scripts on the same box.
Give your agent somewhere permanent to work
Set up takes about ten minutes, and we will help you move your keys and repositories across. Cancel any month.