dagu

dagu runs scheduled jobs and multi-step workflows from a single binary, with no database or message broker required. You define what runs in plain YAML: shell scripts, Docker containers, Kubernetes jobs, SSH commands, HTTP calls, even AI agents over MCP. It ships a web UI for watching runs and digging into logs, and the core is free under GPLv3. It is cron that grew up, with a dashboard and retry logic.

Self-hosting is about as light as orchestration gets. Drop the binary on a server, point it at a directory of YAML files, and you have a scheduler. State lives in files, so there is no Postgres or Redis to babysit. Scaling to distributed workers across machines adds a coordinator and some moving parts, but for most teams a single instance handles real work without much care and feeding.

Solo developers and small teams replacing a pile of crontab entries will love this, and the free tier covers the actual job running. SSO, RBAC, and audit logging sit behind a paid self-host license, with a managed Dagu Cloud if you would rather not run it yourself. Large teams that need access control will hit the paywall, and that is the fair kind: auth is worth paying for.

The catch is that YAML workflows get unwieldy fast. Simple pipelines stay readable, but complex branching logic starts to fight you. If your orchestration needs are heavy, look at Temporal or Airflow before you outgrow this.