daguA lightweight workflow engine built the way it should be: declarative, file-based, self-contained, air-gapped ready. One binary that scales from laptop to distributed cluster. Used as a sovereign AI-agent orchestration infrastructure.

dagu

A lightweight, single-binary workflow engine that runs Directed Acyclic Graphs (DAGs) defined in YAML. It requires no database, message broker, or language-specific runtime, making it perfect for air-gapped environments and scaling from laptops to distributed clusters.

Key Features

• File-Based & Declarative: Workflows defined in simple YAML files with no database dependencies
• Single Binary Deployment: Runs as one self-contained executable with optional distributed worker support
• Built-in Web UI: Cockpit dashboard for monitoring, managing, and debugging workflow runs
• Durable Execution: Automatic retries, cron scheduling, catchup recovery, and lifecycle event handlers
• Enterprise Ready: RBAC, user management, workspaces, REST API, webhooks, and email notifications
• Universal Task Support: Execute shell scripts, Python, Docker containers, HTTP calls, and Kubernetes jobs

Use Cases

• Legacy Script Management: Bring visibility and control to scattered crontabs and shell scripts across servers
• ETL & Data Pipelines: Orchestrate database queries, file transfers, transformations, and validation workflows
• Infrastructure Automation: Coordinate backups, deployments, patching, and server lifecycle operations
• Media & Compute Jobs: Distribute ffmpeg, image processing, and heavy computational workloads across workers
• AI Agent Orchestration: Use AI agents to manage, write, and repair workflows in plain YAML format
• Container Workflows: Compose multi-step pipelines mixing Docker images, Kubernetes jobs, and shell commands

Who Is It For

Dagu targets DevOps engineers, data engineers, infrastructure teams, and organizations seeking a lightweight, self-contained alternative to heavy orchestration platforms. It's ideal for teams running scripts and jobs on-premises or in air-gapped environments who want observability without operational complexity.