Your first deploy, end to end.
A walk through the ninety seconds between git push and a live URL — with nothing hidden, and every step reproducible from the terminal.
Værft is built around a single idea: the distance between your editor and production should never exceed the length of one command. The rest of this page walks through the four steps most teams take on their first day — installing the CLI, linking a project, running a preview, and rolling out to production. If you already have an account, skip to step two.
Install the CLI
The varft binary is a self-contained Go program. We publish it on Homebrew, npm,
and as a shell installer. Pick the one that fits your toolchain.
# install the CLI brew install varft # authenticate in your browser varft login # create a new project from this directory varft init nordlys-api # deploy to production varft deploy --prod
Link your project
From the root of your repository, run varft init.
The CLI auto-detects your stack (Node, Deno, Python, Go, or Rust), finds your build and start commands,
and writes a minimal varft.toml you can commit alongside your code.
Every subsequent varft deploy runs these five stages. The
output is streamed line-by-line; the CLI exits non-zero if any stage fails. No hidden cache
tricks, no arbitrary wait times.
Useful commands
Eight commands cover the day-to-day. Everything else is a flag on one of them.
Run varft --help for the exhaustive list.
| Command | Does | Example |
|---|---|---|
| varft init | Detect stack, write varft.toml | varft init . |
| varft deploy | Build and roll out a revision | --prod / --region fra1 |
| varft logs | Tail logs across regions | --service web -f |
| varft psql | Open a psql shell to your database | --env preview |
| varft rollback | Revert to a previous revision | --to 48fe2c9 |
| varft env | Read and write environment variables | set DATABASE_URL=@postgres |
| varft dev | Run the same runtime locally | --with postgres,redis |
| varft whoami | Show the active account and team | --org nordlys |
What happens on merge
When a pull request merges into main,
Værft rebuilds the revision on clean infrastructure, promotes it through three canary regions,
and watches p50 latency and 5xx rates for seventy seconds before fanning the rollout out to the
remainder of your configured fleet. If any health check fails, the rollout pauses automatically
and you're notified in Slack.
You can promote manually, pause, or roll back at any time from the dashboard or the CLI — see Preview environments and Rollbacks.