Dagu

Orchestration is rarely the job people signed up for, and Dagu is built around that fact. It ships as a single binary with a built-in web UI, needs no external database and no message broker, and runs on Linux, macOS and Windows.

Workflows are directed acyclic graphs written in declarative YAML that sits next to your scripts. Structure stays configuration rather than code, so nothing imports the engine: delete the YAML and the scripts run exactly as they did before.

• Step types: shell commands, Docker containers, Kubernetes Jobs and remote commands over SSH run without modification.
• Scheduling: cron syntax with timezones, overlap policies and catch-up windows.
• Composition: reusable sub-DAGs and concurrency controls for work that runs in parallel.
• Operations: logs, run history, retries, notifications, human tasks and webhook triggers in one place.
• MCP server: assistants can inspect workflows and runs, maintain wiki pages, apply changes and control runs.

State lives in local files, so a single machine can carry thousands of workflow runs per day, queues and resource limits decide how many execute at once, and workers spread execution across machines when one node stops being enough.

Teams point it at ETL pipelines, interdependent legacy scripts, media conversion, server automation over SSH, container and Kubernetes work, and self-service runbooks that non-engineering colleagues can trigger themselves. Installers cover a shell script, Homebrew, npm, Docker and a Helm chart.