A VPS control panel for Node, Python and Docker apps.
You want to push code and have it run, without giving up the server. RespireCloud deploys apps from Git, runs containers behind a strict validator, and exposes every action as an API call and a CLI command. It is in early access.
Deploy and run.
-
Git deploys with rollback
30 framework presets, deploys from Git and a rollback to an earlier release. Workers and cron jobs belong to the app.
-
Runtimes per server
Language runtimes are installed per server through mise with pinned tools, and PHP versions with extensions are managed from the software page.
-
Docker and Compose
A Dockerfile or Compose deploy per account. Your compose file is parsed and rewritten from an allow-list; privileged mode, host networking, devices, extra capabilities and Docker socket mounts are refused, each with the offending line.
-
A catalogue of apps
A signed catalogue of ten apps, among them Gitea, n8n, Nextcloud, Uptime Kuma, Plausible and Ghost. A tampered catalogue is refused.
-
A respirecloud.yaml
Put hooks and deploy settings in the repository.
Everything is an action.
-
An API per action
One endpoint per action from the module registry, dry run plans, session and token authentication, one error shape, server sent events and a per caller OpenAPI document.
-
A CLI
respirecloudctl uses the same action registry with friendly verbs, --dry-run plans and completion.
-
Terminal and editor
A browser terminal and a code editor per account, on their own origin.
-
Previewed changes
Every change is a plan with the exact difference, and can be undone.
Servers, clusters and clouds.
These areas follow the flags in the availability file. While a flag is off, the page says so.
-
Hetzner, DigitalOcean and others
Coming in 1.1: create servers at your cloud provider from the panel; they join your cluster by themselves. Today you install RespireCloud on a server you made at the provider. It runs on any Linux server there, including Hetzner and OVH.
-
Kubernetes
Coming in 1.1: connect a cluster or run k3s on your nodes, deploy from YAML with a dry run or from a Helm chart.
-
Load balancers
HAProxy pools with health checks in the cluster, today. Coming in 1.1: your cloud's managed load balancers on the same page.
What RespireCloud does not have yet.
Read this before you plan.
-
Not a PaaS
RespireCloud runs on your servers. It does not provide servers, build pipelines for many teams or scale to zero.
-
No price yet
Prices are not set. A licence counts hosting accounts, resellers, servers and CDN edges, and there is a free 14 day trial. See the pricing page for what is decided.
-
A public release
RespireCloud is in early access. There is no public download, the production signing keys are not made and prices are not set. One install has run on a real server; everything else ran in our lab.
-
Proof on real backups
The importers have run end to end only on archives we built to match each format. A backup taken from a production server has not been imported yet, so expect differences.
-
A second web server
Sites are served by nginx. Apache .htaccess rules, PHP ini overrides and redirects are not translated by the importer: a site that depends on them needs its rules rewritten.
-
Recovery to a moment
A database restore goes back to the time of a scheduled dump. Recovery to a chosen moment between backups is not built. The database tool (Adminer) is built; it covers MariaDB, MySQL and PostgreSQL on the server itself.
-
An independent security test
An external penetration test has not started.
Questions
- Does it run Node.js and Python apps?
- Yes, as apps deployed from Git with presets, runtimes installed per server, workers and cron jobs, or as containers.
- Can I use it on a Hetzner or OVH server?
- Yes, on any server that runs one of the supported Linux systems. Creating servers from inside the panel at Hetzner, DigitalOcean, Vultr and Linode is part of the Cloud area.
- Can I script it?
- Every action has an API endpoint and a CLI command, with dry runs, scoped tokens and idempotent replay.
- How fast is a Node app on it?
- A small Node app served about 7,600 requests a second at 1000 visitors with the 95th percentile at 153 ms on a 4 vCPU server in our measurement. Details and method are on the performance page.
Read next
Building on your own servers?
Ask for early access and tell us what you run.