Scheduling

Setting up a scheduled Loop

A Loop on the Free plan runs at most once a day, and a run that would start sooner is skipped.

1. Go to Console → Loops and choose Create Loop
2. Select the trigger type Scheduled and give the Loop a name
3. Choose a frequency: One time, Every hour, Daily, Weekly, Monthly, or Custom (cron), and set the time
4. Under Instructions, write what the agent should do on each run (Generate with AI drafts it from the Loop’s name)
5. Select an Internal agent (scheduled Loops require Internal agents)
6. Choose Create Loop. The Loop is on straight away

Note

Scheduled Loops do not receive event payloads. Your Instructions are sent to the agent as its task, and a scheduled Loop can’t be created without them. Template variables like {{date.today}} are not supported — write the task in plain language instead.

---

Common schedule examples

Use case	Cron
Every weekday at 9:00 AM	0 9 * * 1-5
Every Monday at 8:00 AM	0 8 * * 1
Every hour	0 * * * *
1st of each month, 9 AM	0 9 1 * *
Every 15 minutes (the most often a custom schedule can run)	*/15 * * * *

---

Writing scheduled instructions

Since there is no user input, make instructions self-contained:

Your task on each run:

1. Pull GitHub issues opened in the last 7 days from acme/app
2. Group them by label (bug, feature, question)
3. Count open vs. closed
4. Write a brief weekly summary
5. Post the summary to #engineering on Slack

---

Times are in UTC

A Loop you create in the Console runs on UTC: the dialog labels its times Time (UTC), and cron expressions are read in UTC too. Convert from your local time when you set it.

A recurring Loop (every hour, daily, weekly or monthly) may start a few minutes after its set time, so that many Loops set for the same minute don’t all start at once. The delay is fixed for each Loop and is never more than 15 minutes. One-time and custom cron Loops start on time.

---

Monitoring scheduled runs

View these runs in Runs; open one to see that it was Scheduled. When the last run failed, the Loop’s row says Last run failed with the error. Failed runs show the error message from the agent run — check credits, agent instructions, and integration connections if a run fails repeatedly.