Yard runs all your apps on one Linux VM that you own and can read. Each app is a Podman container. Caddy holds port 443 and the certificates. A push to the repo deploys the app. An app is whatever you already have: a repo in any language, an image that someone publishes, or a compose file.
# Your own code, in any language: run it in the repo's checkoutyard app new --caddy app.example.com # A published image: a password manageryard app new --app vault --image docker.io/vaultwarden/server:1.37.4 --caddy vault.example.com # A compose file: a status pageyard app new --compose-url https://raw.githubusercontent.com/stariodev/yard/main/examples/uptime-kuma.compose.yml \ --caddy status.example.comThere is no registry, no CI service, no database, and no dashboard. Yard is the small program that connects tools you already trust, and these docs show you each joint: what happens, and why it happens that way.
Yard grew up next to Stario, the Python framework on this site, so a Stario app needs the least setup of all: no Containerfile and no settings. Yard reads its pyproject.toml and uv.lock and writes the rest. See what it reads.
The docs read as a short book in four parts. Start gives you the idea and one app from your laptop to a live address. Use covers the jobs: a host, a service someone else wrote, a compose file of your own, and the commands of an ordinary day. How it works opens each part of the machine, one chapter each. Reference is for lookups.
Start
- The whole idea — why Yard exists, what each tool does, and what Yard leaves out.
- From laptop to live — one app from a folder to an HTTPS address, then a change, a setting, and a mistake.
- One push, start to finish — a single commit followed from your laptop to a live container.
Use
- Set up a host — a fresh VM to a working Yard in one script and a few commands.
- Self-host a service — Vaultwarden from an image and Umami from a compose file, with no repo.
- Healthchecks — the one line that lets a broken version go back by itself.
- Your own compose file — workers, databases, healthchecks, and apps in any language.
- Day to day — logs, settings, domains, rollbacks, and cleanup.
How it works
- The host — the VM, the yard user, one public port, and the files in /srv/yard.
- Podman and containers — rootless containers, one network, names that Caddy dials.
- What an app runs — one source per app, and how Yard reads a repo.
- Deploys — the check, the deploy container, the healthcheck gate, and the record.
- Caddy and HTTPS — one small Caddyfile per app, and certificates nobody thinks about.
- Settings and data — the env file, the data directory, and backups.
- Logs and events — one journal that outlives every container.
- Forges and webhooks — GitHub and Cursor Origin apps, tokens, and webhooks that only say "go look".
- Inside Yard — the program itself, and how it upgrades itself.
The command is yard; the PyPI package is staryard.
