State you can inspect
Desired state lives in a database and a reconciler records a status condition with a reason after every pass, so you can see why something is not ready.
Kamal alternative
Kamal is deliberately minimal: no server component, just SSH from your machine or CI. Levelrail keeps readiness-gated cutover and adds a dashboard, history, rollback and observability.
curl -fsSL https://levelrail.com/install.sh | sudo sh
Desired state lives in a database and a reconciler records a status condition with a reason after every pass, so you can see why something is not ready.
Every deploy is recorded with its pinned image. Rolling back is switching to an image that is still there.
Servers run an agent that dials out to the control plane, so CI talks to the API with a scoped token instead of holding SSH keys.
Kamal is a sharp tool for people who want the fewest moving parts. These are the design differences.
One script on a Linux server.
Use the CLI or the dashboard and watch the readiness gate.
Use a scoped API token from GitHub Actions or any pipeline.
Yes, by design: it is a control plane with a database and a dashboard, while Kamal has no server component. The control plane idles at a small footprint, which the case studies page measures.
Yes. CI calls the API or CLI with a scoped token. See the GitHub Actions guide.
No. You point Levelrail at the same images and recreate configuration.
Yes. Apache 2.0, one edition.
Try it on one server beside your current setup.