From laguagu-claude-code-nextjs-skills
Manage Hetzner Cloud infrastructure via the `hcloud` CLI: servers, networks, firewalls, load balancers, volumes, DNS, SSH keys, IPs, snapshots, certificates, placement groups, and storage boxes. Activates on mentions of Hetzner, hcloud, VPS, or location codes.
How this skill is triggered — by the user, by Claude, or both
Slash command
/laguagu-claude-code-nextjs-skills:hetzner-cloudThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Operate Hetzner Cloud through the `hcloud` CLI. When in doubt about server types, prices, locations, images, or flag syntax, run the discovery command rather than rely on this file or pretraining — the CLI is current.
Operate Hetzner Cloud through the hcloud CLI. When in doubt about server types, prices, locations, images, or flag syntax, run the discovery command rather than rely on this file or pretraining — the CLI is current.
Generate a token in Hetzner Console → project → Security → API Tokens, then either:
hcloud context create <name> # interactive: stored in cli.toml, persists across shells
export HCLOUD_TOKEN="…" # CI/scripts: most commands read this automatically
hcloud context create is the one common case that still prompts even with the env var set — pass --token-from-env to skip the prompt in non-interactive sessions.
Health check: hcloud datacenter list.
hcloud manages everything inside Hetzner Cloud: servers, networks, firewalls, load balancers, volumes, DNS zones, SSH keys, primary/floating IPs, snapshots, certificates, placement groups, storage boxes.
It does not manage:
Discovery beats memorization. Pretraining ages; the CLI is current.
| Question | Command |
|---|---|
| Server types and prices | hcloud server-type list |
| Locations and datacenters | hcloud location list |
| OS images | hcloud image list --type system (add --architecture arm for ARM) |
| Command/flag details | hcloud <resource> <verb> --help |
| Everything in this project at a glance | hcloud all list |
Structured output for scripts: --output json | jq …, --output columns=id,name,status, --output format='{{.PublicNet.IPv4.IP}}'.
Single commands are well-covered by --help. The chains worth pinning down:
Server with SSH key + firewall, ready to log in — key and firewall must exist before server create references them:
hcloud ssh-key create --name <key> --public-key-from-file ~/.ssh/id_ed25519.pub
hcloud firewall create --name <fw> --rules-file rules.json
hcloud server create --name <srv> --type <type> --image ubuntu-24.04 \
--location hel1 --ssh-key <key> --firewall <fw>
hcloud server ssh <srv>
Replace firewall rules atomically (don't drift via repeated add-rule):
hcloud firewall replace-rules --rules-file rules.json <fw>
Public IPs get recycled. A new server may inherit a previously-used IP and SSH refuses with REMOTE HOST IDENTIFICATION HAS CHANGED!. After deleting a server, ssh-keygen -R <ip> and let ssh-keyscan re-add the new host key.
Context vs HCLOUD_TOKEN. Context (in user-config cli.toml) persists across shells — preferred for interactive work. Env var is preferred for non-interactive use, paired with --token-from-env. A token sitting in a .env file is not automatically exported — shells must source it.
Volumes are location-pinned. A volume in fsn1 cannot attach to a server in hel1. Match --location at creation.
hcloud zone is only half of DNS. The registrar's NS records must point to Hetzner's nameservers before queries resolve. Without that, the zone is just an internal database.
Match location to where users are; pretraining defaults to fsn1. Run hcloud location list and pick the closest. Private networks can't span network zones, only locations within one: eu-central (fsn1/hel1/nbg1), us-east (ash), us-west (hil), ap-southeast (sin).
Don't quote prices or server-type specs from memory. Always verify with hcloud server-type list before committing — names, specs, and pricing shift.
Deletion is immediate and unrecoverable. No undo on server delete or volume delete. describe first; create a snapshot (hcloud server create-image --type snapshot <srv>) beforehand if you might want the server back.
Hetzner Cloud ≠ Hetzner Robot. If the user mentions dedicated servers, server auctions, or the Robot dashboard — that's the other product. Surface the distinction; don't try to manage it here.
npx claudepluginhub laguagu/claude-code-nextjs-skillsManages Hetzner Cloud infrastructure via CLI: servers, volumes, firewalls, networks, DNS, snapshots. Enforces safe operations with confirmation prompts and snapshot suggestions.
Guides hcloud CLI for Hetzner Cloud: provision servers, networks, firewalls, load balancers, volumes with decision trees and commands.
Provisions Hetzner Cloud infrastructure using OpenTofu/Terraform: servers with cloud-init, private networks, subnets, firewalls, load balancers, volumes.