Closed Networks

Dagu schedules and runs workflows without internet access. It needs no external database, broker, or license server. Some optional features do need a connection, and this page lists them.

Install

On a machine with internet access, download the archive and checksums.txt for your platform from https://github.com/dagucloud/dagu/releases and verify them. Then move the archive into the closed network and install the binary:

tar -xzf dagu_2.26.0_linux_amd64.tar.gz dagu
sudo install -m 0755 dagu /usr/local/bin/dagu
dagu version

For containers, copy the image to your own registry or load it directly:

docker pull ghcr.io/dagucloud/dagu:2.26.0
docker save -o dagu-2.26.0.tar ghcr.io/dagucloud/dagu:2.26.0

# inside the closed network
docker load -i dagu-2.26.0.tar

For Kubernetes, download the chart with helm pull dagu/dagu, push the image to your registry, and set image.repository and image.tag. See Images and private registries.

Configure

Turn off the update check, which otherwise tries to reach GitHub:

check_updates: false

License

Servers that cannot reach Dagu Console use an offline license.

1. From a machine with internet access, create one with Add offline server on the console's Instances page.
2. Install it with DAGU_LICENSE or DAGU_LICENSE_FILE, or save it as data/license/license.jwt.

An offline server takes a license slot until its license is revoked. Paid features keep working for 14 days after a license expires, so replace it before then.

An active and standby pair that shares a data directory needs one license.

Private Control Plane

Organizations that need central monitoring and alerts inside a closed network can have the control plane deployed in their own network. This is available on request under an Enterprise agreement, with the scope and schedule agreed per engagement. Contact company@descarty.com.

Your servers report to it instead of Dagu Cloud. Because it runs separately from them, it can also alert when a server stops reporting.

What needs a connection

Feature	Needs	Without it
Monitoring in Dagu Console, including alerts for failed workflows, missed deadlines, and servers that stop reporting	console.dagu.sh	Use a private control plane, or see below.
Public webhook URLs	console.dagu.sh, relay.dagu.sh	Use the local webhook endpoint from inside your network.
Remote access from Dagu Cloud	console.dagu.sh, relay.dagu.sh	Connect MCP clients to the server directly.
dagu upgrade	GitHub	Replace the binary by hand.
tools in a DAG	GitHub, for the aqua registry and downloads	Install the commands on the host or in the image, and remove tools from the DAG.
Dagu Actions	GitHub	Use built-in step types and your own commands.
AI steps	The model provider	Point them at a model endpoint inside your network.

Alerts without Dagu Console

Deadline, late-start, and server-stopped alerts come from Dagu Console. Without a private control plane, build the alerts you need from what runs locally:

• Failed workflows: a handler_on.failure step, or email through your internal mail server.
• Scheduler health: the dagu_scheduler_instances and other Prometheus metrics, and GET /health on the scheduler.
• A job that never started: a separate check that compares the DAG's last run against when it was due, using dagu history or the API.