Setup

Getting Started with Kamal

You are an expert in deploying applications with Kamal. Your goal is to take a user from nothing to a running first deploy: install the tool, generate the config, fill in the values, and run the first kamal setup against their servers.

Before You Start

Check what already exists before asking questions:

• If config/deploy.yml is present, read it first and work from the values the user already has instead of starting from scratch.
• If .kamal/secrets is present, read it to see which secrets are already wired up (do not echo secret values back to the user).

Then confirm the essentials you will need:

1. Servers — the IP addresses or hostnames Kamal will deploy to.
2. Registry — which Docker registry the image is pushed to, and the username/password for it.
3. Image name — what the built image will be called.
4. SSH access — kamal setup connects over SSH (root by default, authenticated by your SSH key).
5. Committed code — Kamal builds from a git archive of HEAD, not your working tree; commit every file the image needs before deploying (see the preflight note in Step 5).

Step 1: Install Kamal

If you have a Ruby environment available, install Kamal globally:

gem install kamal

If you do not have Ruby installed, you can run Kamal from a Docker container instead. On macOS:

alias kamal='docker run -it --rm -v "${PWD}:/workdir" -v "/run/host-services/ssh-auth.sock:/run/host-services/ssh-auth.sock" -e SSH_AUTH_SOCK="/run/host-services/ssh-auth.sock" -v /var/run/docker.sock:/var/run/docker.sock ghcr.io/basecamp/kamal:latest'

The dockerized path has limitations (the secret adapters and host environment variables are not available in the container). For the Linux alias and the full list of limitations, see references/dockerized-install.md.

Step 2: Initialize the Project

Inside your app directory, run:

kamal init

This creates the files you need to deploy:

Created configuration file in config/deploy.yml
Created .kamal/secrets file
Created sample hooks in .kamal/hooks

No Dockerfile yet? Kamal builds the standard Dockerfile at your project root. Rails generates one with rails new; most other frameworks (Node/Next.js, Django, Go) do not — write one before kamal setup, or build without one via buildpacks (see the build skill). A good shape is multi-stage: a deps stage that installs dependencies, a build stage that compiles the app, and a slim runner stage that copies only the built output. Common footguns: native modules often ship prebuilt binaries, so try building without a C toolchain before adding one; the framework's build step usually doesn't need runtime secrets or database access — verify by building with them absent instead of threading them through as build args; and run migrations from an ENTRYPOINT script at container start, not at build time — volumes aren't mounted during the image build.

Step 3: Write config/deploy.yml

Edit the generated config/deploy.yml. A first deploy can be as simple as this:

service: hey
image: 37s/hey
servers:
  - 192.168.0.1
  - 192.168.0.2
registry:
  username: registry-user-name
  password:
    - KAMAL_REGISTRY_PASSWORD
builder:
  arch: amd64
env:
  secret:
    - RAILS_MASTER_KEY

What each key does:

The default registry is Docker Hub; you can change it with registry/server. For every available configuration key, see the config skill.

App ships its own stack definition? If the project has a production docker-compose.yml, Helm chart, or upstream install docs, treat it as the contract: map every service it runs to a Kamal accessory, an external service, or an explicit user-approved omission. Never silently drop or swap a component because the host looks too small — surface the constraint and let the user choose before deploying. See "Mirroring an existing stack" in the accessories skill.

Persisting data across deploys

Every deploy replaces the app container — anything written to the container's own filesystem (a SQLite database, uploaded files, anything else meant to persist) is discarded with the old container. Mount a named volume for any path that must survive:

volumes:
  - "myapp_data:/app/data"

Point your app's writable paths at the mounted directory (e.g. DATABASE_URL=file:/app/data/production.db), and chown that path to the container's runtime user in the Dockerfile — Docker copies ownership from the image into a newly created named volume on first use. Then confirm the data actually survives a redeploy (run kamal deploy again and check) before trusting it — a plain docker restart keeps the same container and proves nothing.

Step 4: Fill In .kamal/secrets

The registry/password and any env/secret entries above are looked up by name in your environment and in .kamal/secrets. Set the registry password in your environment, then edit .kamal/secrets to read it (plus RAILS_MASTER_KEY for a production Rails app):

# .kamal/secrets
KAMAL_REGISTRY_PASSWORD=$KAMAL_REGISTRY_PASSWORD
RAILS_MASTER_KEY=$(cat config/master.key)

Each line maps a secret name to a value: an environment variable ($KAMAL_REGISTRY_PASSWORD) or the output of a command ($(cat config/master.key)). For reading secrets from 1Password, LastPass, Bitwarden, AWS, GCP, Doppler, and other vaults, see the secrets skill.

Step 5: Bootstrap and Deploy with kamal setup

With the config and secrets in place, run your first deploy:

Preflight — commit first. Kamal builds from a git archive of HEAD, not your working tree. Files that are new or modified but uncommitted — generated or newly added files (config/*.rb, freshly created credentials, view overrides) — are excluded from the image and fail the build with COPY <file>: not found. Commit everything the Dockerfile COPYs before running kamal setup. To build the working tree instead, set context: . (see the build skill).

kamal setup

kamal setup runs everything required to deploy to a fresh host:

1. Install Docker on all servers, if it has permission and it is not already installed.
2. Boot all accessories.
3. Deploy the app (the same work as kamal deploy).

What kamal setup does step by step

Expanded, the first deploy:

1. Connects to the servers over SSH (root by default, authenticated by your SSH key).
2. Installs Docker on any server missing it (using get.docker.com); root access via SSH is needed for this.
3. Logs into the registry both locally and remotely.
4. Builds the image using the standard Dockerfile in the root of the application.
5. Pushes the image to the registry.
6. Pulls the image from the registry onto the servers.
7. Ensures kamal-proxy is running and accepting traffic on ports 80 and 443.
8. Starts a new container with the version of the app that matches the current Git version hash.
9. Tells kamal-proxy to route traffic to the new container once it responds with 200 OK to GET /up.
10. Stops the old container running the previous version of the app.
11. Prunes unused images and stopped containers so servers don't fill up.

All servers are now serving the app on port 80. If you run a single server, you're ready to go. If you run multiple servers, put a load balancer in front of them.

If kamal setup is interrupted (SSH drop, network reset) partway through: the deploy lock may still be held on the server — check with kamal lock status. If the image already built and pushed, resume without rebuilding: kamal lock release, boot any accessories that aren't running yet (kamal accessory boot all; a name-conflict error for one that already booted is harmless), then kamal deploy --skip-push to deploy the already-pushed image. If the failure happened during build or push, just release the lock and rerun kamal setup. For lock details, see the deploy skill.

Bootstrapping Docker separately

kamal setup installs Docker for you when it has permission. To set up Docker on the hosts explicitly — for example before a deploy — run:

kamal server bootstrap

This checks whether Docker is installed and, if not, attempts to install it via get.docker.com.

After the First Deploy

For subsequent deploys, or if your servers already have Docker installed, you do not need setup again — just run:

kamal deploy

kamal deploy builds and deploys the currently checked-out version of the app to all servers. For the full release workflow, see the deploy skill.

Useful Commands

Most-used global options

For the full command list, the kamal server subcommands, and every global option, see references/cli-reference.md. You can also run kamal --help or kamal help [command] at any time.

Related Skills

• config: For the full set of config/deploy.yml keys and defaults.
• deploy: For routine releases with kamal deploy after the first setup.
• secrets: For .kamal/secrets and reading secrets from vault adapters.
• registry: For configuring Docker Hub, ECR, GCP Artifact Registry, and other registries.