Fresh VPS

From an empty VPS to a secure web server, by asking

You rent the server, run one line, and tell your AI what it is for. It sets up a sane, updated, secured base the way an experienced sysadmin would, explains each step, and leaves SSH alone until you confirm your key works.

Works with Claude, Claude Code, ChatGPT and Cursor.

How your AI does it

  1. STEP 1

    It looks before it touches

    server_facts tells your AI the distribution, package manager, RAM, disk and what already listens. If the server is not actually empty, it stops and asks.

  2. STEP 2

    Updates and basics

    Full upgrade, EPEL on Rocky and Alma, and the everyday tools. Package runs happen as background jobs, so a slow mirror does not time out the chat. If the kernel was upgraded it tells you a reboot is due and asks when.

  3. STEP 3

    Swap, time and hostname

    A swap file sized to the RAM when there is none, swappiness turned down, chrony for the clock, and the hostname and time zone you give it.

  4. STEP 4

    Firewall and fail2ban

    firewalld on Rocky and Alma, ufw on Debian and Ubuntu. SSH is allowed first, on whatever port it really uses, then HTTP and HTTPS if a site will run. fail2ban watches SSH.

  5. STEP 5

    Automatic security updates

    dnf-automatic or unattended-upgrades applies security updates on its own from then on.

  6. STEP 6

    SSH, only with you

    Switching SSH to keys only is offered only if a key is installed and you confirm you log in with it, as its own step, after sshd -t. Locking you out is the one outcome it is built to avoid.

Works on

  • Rocky Linux, AlmaLinux and RHEL (dnf)
  • Debian and Ubuntu (apt)
  • Any Linux with systemd, x86_64 or ARM64
  • Hetzner, DigitalOcean, OVH, Vultr, Linode, AWS, Google Cloud, Azure or your own hardware
  • Servers that block inbound ports: the agent only dials out over HTTPS
  • Then Magento, WordPress or any other stack, by asking

Tested on real servers

From our own test runs with Claude Code on fresh Rocky Linux 10 servers.

1 line

to connect a server: the agent installs as a systemd service and shows up online in the dashboard within seconds

0 ports

opened for SudoWhizzy: the agent calls out over HTTPS, and no SSH key or password is stored anywhere

Start with this prompt

Paste it into your AI with SudoWhizzy connected. It fetches the matching playbook and explains its plan before changing anything.

Using SudoWhizzy, set up my empty server as a Magento shop. Before you start, check the server with server_facts and ask me only for what you do not know, explaining why you need each: - the shop's domain (it should already point at this server) - Mage-OS (no Adobe keys needed) or Magento Open Source (I give you my Adobe access keys) - admin email and name, store language, currency and time zone Then follow the server-setup and magento playbooks (get_playbook). Run long steps with start_job. Do not paste passwords into the chat: write them to a root-only file and tell me where. Finish with a short report: versions, URL, admin path, what I should do next.

Set up a fresh server as a Magento shop · all prompts

Which plan

Starter and up, since setting up a server means changing it. Free lets your AI look at a server and tell you what it would do. Pricing.

Questions

Which provider should I use?

Any provider that gives you root on a Linux VPS. For a Magento shop pick at least 4 GB of RAM; for WordPress 2 GB is comfortable. Your AI checks the size first and tells you if it is too small.

Does the agent need a public IP or open port?

No. It dials out to sudowhizzy.com over HTTPS, so it works behind NAT and strict firewalls, and you do not open anything for it.

What if a package mirror stalls?

Your AI retries with shorter timeouts or another mirror. It never switches off package signature checks to get past an error; if something cannot be installed safely, it stops and tells you.

Can I see exactly what it ran?

Yes. Every command, file write and result is in your activity log with its risk level, and /etc is saved before changes so you can roll back.

Safety on every task: destructive actions wait for your approval, /etc is saved before changes, and every action is logged. How it works.

Connect your first server

Start free with one server in read-only mode. Upgrade when you want your AI to fix things.

Get started