Getting started.

Get access

dagweave is a hosted app, and we're bringing people on in batches. Leave an email on the waitlist and we'll get in touch when your batch comes up. Once you're in, sign in with email or with GitHub.

There's nothing to install to start. Open the builder in a browser and you're on the canvas. The connector is the only thing that goes into your infrastructure. That comes later, once you're ready to dispatch runs instead of just designing them.

Your first workflow

1. Open the builder and add a node to the canvas, or pick a catalog item from the node picker.
2. Set its container image and command, or fill in the catalog item's parameters.
3. Add a second node and draw a dependency between them.
4. Pick a kind: WorkflowTemplate, Workflow, or CronWorkflow.
5. Publish the compiled YAML to Git, or dispatch it straight to your enrolled cluster.

The YAML dagweave produces is standard Argo Workflows: the same format kubectl and argo submit already expect.

Import existing YAML

Import a manifest and dagweave lays it out on the canvas: names, containers, entrypoints and kind, intact. Change settings in the UI instead of by hand. What you publish back is a small diff. Point it at a whole GitHub repo, or read definitions straight off a live cluster through the connector.

kind: WorkflowTemplate   # preserved on export
metadata:
  name: my-pipeline       # preserved
spec:
  entrypoint: main         # preserved

Dispatch to your cluster

Connecting a cluster lets the canvas dispatch real runs. An operator enrolls the connector once with a single helm install and points it at your Argo Workflows instance. From then on dagweave dispatches runs, WorkflowTemplates and CronWorkflows to that cluster, and authors trigger them from the app with no cluster access of their own. Enroll as many clusters as you have.

Runs start manually, on a cron, from a signed webhook, from GitHub or from an AWS SNS topic. Phase, per-step status, log references and artifact URLs come back next to the canvas. The failing step shows its error, and you can re-run it in one click. Your own Argo Workflows controller executes every run.

Next steps