Guides
Asking for an automation
How to describe a scheduled or incoming automation, and what the assistant does next.
You don't need to know any code. Describe what should happen, when, and where the result should go.
A good request has three parts#
| Part | Examples |
|---|---|
| What to do | "check this page", "total yesterday's orders", "summarize new issues" |
| When | "every hour", "weekdays at 9 New York time", "when Stripe sends a payment" |
| Where it goes | "email me", "post to #ops in Slack", "only tell me if it changed" |
If you leave something out, the assistant asks. It will also ask for your timezone when a schedule depends on a local hour.
Scheduled automations#
Example prompt: Every hour, check https://status.example.com and Slack #ops if any component isn't operational.
Example prompt: Every Monday at 8am London time, email me last week's new GitHub stars on our repo.
Supported schedules: every N minutes (5, 6, 10, 12, 15, 20 or 30), every hour at a minute, every day at a time, weekdays at a time, one weekday at a time, one day a month (1st to 28th, or the last day) at a time, or, for anything else, a cron expression such as "the second Tuesday of each month at 9". Schedules run at the local time of the timezone you choose. You'll see the next three run times before approving.
Automations triggered by an event#
Example prompt: When a Stripe checkout completes, post the amount and customer email to #sales.
Example prompt: Collect every Typeform submission today and email me one digest at 6pm.
The assistant creates the automation with a trigger link. Paste that link into the sending service. See Trigger links.
What the assistant does#
- Restates the job in a sentence or two and asks only for what's missing.
- Saves a draft. Drafts do nothing on their own.
- Shows the schedule so you can correct the time.
- Lists setup it needs: a connection, a secret, or a trigger link to paste somewhere.
- Tells you what will happen: what will run, how often, and which outside services it will contact.
- Turns it on and tests it. Before running something live it says what effects the run will have. If a run fails, it fixes the script, redeploys and tries again until a run succeeds, then tells you when it runs next. Your AI assistant asks for your permission before each of these actions.
Looking after it#
You can ask about any automation at any time:
Example prompt: Did my status page check run this morning?
Example prompt: Why did the Stripe → Slack automation fail?
Example prompt: Pause the weekly report. / Move the forecast email to 7:30.
Example prompt: Post to #alerts in Slack whenever the Stripe automation fails.
Turning it on, applying changes, running it now, resuming it and sending a test event all go through your AI assistant's permission prompt. Deleting an automation also needs a confirmation: a card in the chat, or a one-time confirm link that works once and expires after 30 minutes. Pausing is immediate.
If something is wrong with ClickRuntime itself rather than your script, the assistant can report it to support for you. Replies appear under Issues in the ClickRuntime dashboard, and the assistant can read them too.
What the assistant won't do#
It declines jobs involving malware, secretly monitoring a person, stealing credentials, getting around another service's access controls or rate limits, moving money, placing trades or making purchases, or collecting health, government ID or payment card data.
When monthly runs are used up#
Runs stop until the next period starts; there are no overage charges. You can still turn on or resume an automation, and it starts running when the period resets. The assistant warns you when an automation is likely to use up the month's runs.