Automations
Link an event to a chain of actions: a schedule, a commit, a finished agent, GitHub, GitLab or a webhook call.
An automation links an event to an action. You say when it fires (a schedule, a commit, an agent that finished, a pull request, a call from another system) and what it does (run a command, send a request to an agent, open a new agent on the Canvas, write to a note, show a notification). Elyra runs it on its own and keeps what happened so you can read it later.
It suits repeated work that does not need you there: reviewing what was opened, running the tests on every commit, telling you when an agent is waiting for you, reacting to a PR opened on GitHub.
Where to find it
In the menu at the corner of the window, choose Automations. The screen has the list on the left, grouped by project, and the details of the selected automation on the right.

The sentence: when, where and what
Every automation reads as a sentence:
When there is a commit in my-project, in my-project, on main, run a command in the terminal, and then notify “Commit in the project”.
| Part of the sentence | What it defines |
|---|---|
| When… | The trigger. See Triggers |
| in (project) | The project the automation works in |
| in (workspace) or in a new workspace | Where the actions run: a workspace that already exists, or a new one on every run |
| do this, and then… | The chain of actions, in order. See Actions |
What the screen shows
- The list, with the filters All, On, Off and Need attention, and a search by name. Each row shows the trigger, the first action and when it last ran.
- The state: on, off, running, failed or needs attention. The switch in the corner turns the automation on and off.
- Needs attention appears, for example, when an event automation lost its trigger because it was saved by an old version of Elyra. Edit it, choose the trigger again, and it runs again.
- What happens, in order: the trigger and each action, in the order they run.
- The history: one row per run, with the status, “X of N actions” and the reason when it did not run. See Running now and reading the history.
The More actions menu (the three dots) has Run now, Duplicate and Delete.
What an automation is, and what it is not
- It runs while Elyra is open. If Elyra is closed at the scheduled time, the run does not happen and is not run later; it is recorded in the history as “not run”. The exception is the Elyra Cloud server without a window, which runs the chain without needing a window (see Webhook).
- It can run an agent, but also a plain command. When there is an agent, the request is plain text and the agent decides how to carry it out, with the tools and permissions it has.
- It stops at the first failure. If an action in the chain fails, the following ones do not run.
- It does not replace supervision. What it does happens without you watching; read the history before trusting it with anything sensitive.
- It works in a project. The work happens in that project’s repository.
Turning off is not deleting
The switch turns the automation off and on. Turning off stops the runs and keeps everything: the trigger, the actions and the history. Deleting is a different operation, and a destructive one (see Editing and deleting).
Command line
All of this can also be done from the terminal, with elyra automations; the triggers,
actions and webhook pages give the commands for each case.

