Reference
API
The panel and the CLI both run on Nebla's REST API, so anything you can do in the browser you can script. The API lives under /v1 on your panel and is described by an OpenAPI 3 spec in the repository.
Authentication
Create a token on a project's API tokens page (see Projects and access) and send it as a bearer token:
curl -H "Authorization: Bearer $NEBLA_TOKEN" https://panel.example.com/v1/servers
A token belongs to one project, so requests need no project ID. A Read token can only make GET requests. A token reaches the project's locations, images, server types, SSH keys, servers, volumes, firewalls and actions. Account, project and admin endpoints need a browser session.
Resources
| Path | What it holds |
|---|---|
/locations, /images, /server_types | What the project can create servers from |
/ssh_keys | The project's SSH keys |
/servers | Servers, with power, rebuild, change type, console and metrics under /servers/{id}/… |
/volumes | Volumes, with attach, detach and resize actions |
/firewalls | Firewalls, with set rules, apply and remove actions |
/actions | Every action in the project |
Objects can be looked up by ID. Lists take filters such as name, status and label_selector.
Create a server
curl -X POST https://panel.example.com/v1/servers \
-H "Authorization: Bearer $NEBLA_TOKEN" \
-H "Content-Type: application/json" \
-d '{"name": "web1", "server_type": "small-2", "image": "debian-13", "location": "home", "ssh_keys": ["laptop"]}'
server_type, image, location and each SSH key take an ID or a name. The response has the new server, the action that creates it, and next_actions that follow it.
Actions
Anything that changes a server, volume or firewall runs in the background as an action. The call returns the action straight away:
{
"action": {
"id": 812,
"command": "start_server",
"status": "running",
"progress": 0,
"resources": [{ "id": 42, "type": "server" }],
"error": null,
"started": "2026-11-21T10:04:12Z",
"finished": null
}
}
Poll GET /v1/actions/{id} until status is success or error. The CLI backs off from one second to five between polls. When an action fails, error has a code and a plain message.
Server actions are POST /v1/servers/{id}/actions/{name}, where the name is poweron, poweroff, shutdown, reboot, reset, rebuild or change_type. A second change to a resource while an action runs on it is refused with 409 and the code locked.
Errors
Errors use HTTP status codes and a JSON body:
{ "error": { "code": "not_found", "message": "not found" } }
Pagination
List endpoints take page (from 1) and per_page (default 25, at most 50), and return the total in meta:
{ "servers": [], "meta": { "pagination": { "page": 1, "per_page": 25, "total_entries": 0 } } }
Label selectors
label_selector takes comma-separated terms that must all match: key=value, key!=value, key (the label exists) and !key (it doesn't). For example role=web,env!=prod.
Rate limits
Each token can make 3600 requests an hour, with bursts of up to 100. Every response carries RateLimit-Limit, RateLimit-Remaining and RateLimit-Reset. Past the limit, the panel answers 429 with the code rate_limit_exceeded and a Retry-After header.
Health
GET /v1/healthz needs no token and answers 200 while the panel is up. Use it for load balancer and uptime checks.