Hetzner

Nuphos authenticates to Hetzner Cloud with a project-scoped API token, which lets the agent inspect and operate the resources in that project.

​

Connect

​

Create an API token

In the Hetzner Cloud Console, open your project and go to Security → API Tokens. Create a token with Read & Write access. Hetzner tokens are scoped to a single project — to work with several projects, connect one token per project.

​

Connect in Nuphos

In Connectors → Connect → Hetzner Cloud, enter: Nuphos verifies the token against the Hetzner API before saving; an invalid or revoked token is rejected. The token is stored encrypted.

​

In the Nuphos app

Hetzner work happens primarily through the agent — there’s no dedicated browsing surface for Hetzner resources yet. The binding itself lives in Connectors, where an Administrator can manage its allow-list or remove it.

​

What the agent can do for you

The agent operates Hetzner Cloud via the hcloud CLI in its sandbox — servers, volumes, networks, firewalls, load balancers, floating IPs, and images. Ask it things like:

• “What’s running in the Hetzner project, and what does it cost?” — it lists servers and resources and summarizes.
• “Provision a new server for the staging runner” — it proposes a plan, and creates the server once you approve.
• “Open port 443 on the web firewall” — a change, so it pauses for your yes.

​

Permissions & authorization

Two boundaries govern what happens through this connector:

• Hard boundary — who can use it. Only team Administrators can bind or remove the token, and they can restrict it to specific members with a member allow-list. Within a session the agent only holds credentials you explicitly selected, re-checked on every request.
• Soft boundary — what the agent may do with it. Reads are free-form; changes (create, delete, resize, reboot, …) always go through plan approvals, and Auto-authorization governs each hcloud command — read-only ones run on their own, destructive ones pause for a human, and your standing rules cover the middle.

The outer limit is the token itself: it’s scoped to one project, so the agent can never reach beyond that project no matter what it’s asked.

​

Good to know