Automation: actions, triggers and scheduled tasks
Connect your app to other services, react to events automatically and let repetitive work run on its own — without writing code.
For safety your app cannot reach out by itself: an API key placed in the page is a leaked key. Work that needs the outside world happens in ACTIONS — on the server, on your behalf. And an action doesn't have to be one-off: you can schedule it.
How an action comes to life
Setting up your first action
- 1Dashboard → Connect → Actions
In the "New action" box, describe the job in plain words: which service to call, what to store, who to notify.
- 2Add the key if one is needed
If the recipe needs a secret key you enter it once in the Secret keys section on the same page. The value is stored encrypted and never read back — nobody (including you) sees it again.
- 3Read the recipe
The steps and allowed addresses are listed. If something doesn't make sense, don't enable it; discard and re-describe the job differently.
- 4Test it
The "Test" button runs it with the real keys even while it's off, and shows what each step did. The right order: test first, then enable.
- 5Attach it to an event if you like
In the Triggers section on the same page you say "when a record is added / updated / removed in this table, run this action". Notify on a new appointment, tell the stock service when an order is deleted — without you pressing anything.
- 6Schedule it if needed
In Scheduled tasks on the same page you pick a frequency and time. Your timezone is respected, and you can pause a task without deleting it.
What it's for, and what it isn't
- Pulling information from another system
- Receiving a payment or form notification
- A morning summary, nightly sync, weekly report
- Reacting on its own when a record is added/removed (a trigger)
- Any job whose key must NOT be visible in the browser
- Looping over thousands of records in one run (there are no loops)
- Long-running work — runtime is deliberately bounded
- Ordinary screen logic (that's the builder's job)
Try it yourself: a harmless action
See the whole flow with an action that needs no key.
- 1In Connect → Actions write: "fetch a short sentence from api.github.com/zen and return it".
- 2Read the recipe: you should see one request step and one allowed address.
- 3Press "Test" — see the request's outcome in the run log.
- 4Save and enable; then add a daily 09:00 task for the same action.
- 5Pause the task, then delete it — make sure you leave things clean.
You've seen the describe → test → enable → schedule loop end to end; in a real integration the only difference is entering a key.
No. A well-built screen still works with the action absent: the capability is hidden or a calm note is shown. The dashboard shows the action's last run status and error.
No — by design. The value is stored encrypted and never read back; to change it you write a new one over it. That's the correct behaviour for token management.
No. Writes made by an action itself do not count as trigger events — the system deliberately keeps action-originated writes out of the event queue. And a triggered action still respects the hourly run cap.
Key takeaways
- An action is work the server does on your behalf; you don't write code, you approve a recipe.
- Test before enabling — with the real keys, while it's still off.
- A scheduled task needs no cron knowledge: frequency, time, your own timezone.
- A trigger attaches an approved action to a data event: it runs on its own when a record is added/changed/removed, with loop protection built in.
Related topics
Still stuck? Ask the assistant in the corner, or browse the FAQ.