$DAGU is live on @ponsdotfamily
CA : 0x15b70c26c4d3d5ff394e11daefda8ffa552d1c4e
Dagu is GPLv3 and free forever. Unlimited servers, unlimited workers.
If it saved you from one more fragile crontab, a star helps other devs find it:
https://t.co/66Wcmg3u89
The boring stuff that keeps 2 a.m. jobs alive, by its real name:
retry_policy · catchup_window · overlap_policy · CRON_TZ · lifecycle handlers · zombie detection
All in plain YAML. All in the free, open-source binary.
Same binary, three ways to run it:
1. Single server: dagu start-all, done
2. Temporary workers: one per run, destroyed after
3. Coordinator + workers over gRPC, routed by labels like gpu=true
Start small. Scale out without switching tools.
Your AI agent can now run your workflows.
Dagu ships a built-in MCP server. Point Claude Code, Cursor, Codex, Gemini CLI or any MCP client at /mcp and it can:
• create & edit workflows
• start, stop, retry runs
• read logs
Scoped by API-key roles.
https://t.co/PfR4VuhnZl
Honest question for ops & data folks:
What's the scariest job still running on a bare crontab in your stack? 👇
Bonus points if nobody remembers who wrote it.
From zero to your first workflow in under a minute ⏱
1. curl the installer
2. write hello.yaml
3. dagu start hello.yaml
4. dagu start-all → Web UI on :8080
macOS, Linux, Windows, Docker, Helm, npm.
https://t.co/ngIpaAh1Vf
Status, history, logs and manual reruns all show up in the Web UI.
Start with ONE nightly job, not a platform migration.
Full guide: https://t.co/PuMEiPY7dV
Cron has run the world's scripts for 50 years.
It's also why nobody noticed the nightly job failed 3 days ago.
How to modernize cron jobs without rewriting a single script 🧵
Then add what production needs, one key at a time:
• retry_policy per step (limit, interval_sec)
• catchup_window to replay missed runs
• overlap_policy: skip, all or latest
• CRON_TZ=Asia/Jakarta for timezones
Most workflow engines make you run a platform before you run a single job: scheduler, webserver, metadata DB, broker, executors.
Dagu is one binary. State lives in local files, so it even works air-gapped.
Install. Write YAML. Run.
https://t.co/CzrUqsUYsu
Your scripts already work. Running them shouldn't need a platform team.
Dagu = one YAML file + one open-source binary:
→ schedules & retries
→ human approvals
→ logs & a Web UI
No database. No framework.
Your code is the engine. Dagu is the cockpit.
https://t.co/CzrUqsUYsu