dagucloud/dagu
Local-first workflow engine for teams whose main work isn't orchestration. Declarative YAML over your scripts, SSH commands, containers, and AI agents; keep workflows separate from business logic. One binary, no database. Airflow alternative.
What it solves
Dagu 是一個本地優先的工作流程引擎，專為需要自動化操作卻不想承擔管理複雜編排平台負擔的團隊設計。它取代了不透明的 cron 工作與像 Airflow 或 Temporal 這類龐大的框架，讓使用者能將既有腳本、容器與作業手冊轉換為具排程、重試與視覺化監控的可投入生產的管線。
How it works
工作流程以有向無環圖（DAG）的宣告式 YAML 格式定義。Dagu 以單一二進位檔運行，內建 Web UI，且不需要外部資料庫或訊息代理。它可以在單機上以獨立模式執行，或透過協調者與 gRPC 分散式工作者進行擴展。支援執行 Shell 指令、Docker 容器、Kubernetes Job，以及透過 SSH 的遠端指令，並透過內建的 Model Context Protocol（MCP）伺服器與特定的 harness 來整合 AI 代理的程式碼 CLI。
Who it’s for
此工具為運維團隊、資料工程師與開發者而建，適合已有自動化腳本，想在不重寫成特定 SDK 或管理複雜基礎設施的情況下，加入可觀測性、相依管理與排程功能。
Highlights
- Single Binary Installation: Self-contained deployment with no external DBMS required.
- AI Agent Integration: Built-in MCP server and
harness.runaction for integrating AI coding agents into workflows. - Language Agnostic: Works with any existing script or tool via shell, Docker, or Kubernetes.
- Human-in-the-loop: Built-in approval steps for manual intervention within automated pipelines.
- Distributed Scaling: Ability to spread execution across a fleet of workers using labels.
- Secret Management: Secure log masking to prevent credentials from leaking into the UI or logs.