Skip to main content

Creating Scheduled Agents

Set an Agent to run on its own — daily, weekly, or once at a specific time — and have the output land in your inbox.

What's a Scheduled Agent?​

A Scheduled Agent is a regular Agent set to run automatically on a schedule. Same Agent, same prompt, same variables — it just fires on its own and emails the output to the people you choose.

This is where Agents stop being a "thing you run" and start being a "thing that runs for you."

For the full picture of what Agents can do — variables, MCP actions, triggers, sharing — see the Agents page. This walkthrough focuses just on the schedule itself.

Before You Start​

Pick or create the Agent you want to schedule. A few things make an Agent worth scheduling:

  • The output is useful on a regular cadence — a daily brief, a weekly summary, a Monday-morning prep doc.
  • The inputs don't change much run to run — schedules use the variable values you've predefined on the Agent. Update the variable once when something changes; the next run picks it up.
  • You want the output in your inbox or someone else's, not just sitting in chat history.

Setting Up the Schedule​

Open the Agent. Scroll to the Schedules section and click Add.

Fill in the fields:

Prompt Variables​

Set the values the Agent uses each time this schedule runs. These are independent of any values used in on-demand runs — schedule its own inputs.

Twin to ask​

The Twin the Agent queries when the schedule fires. Defaults to your own Twin. Change it if this scheduled run should pull from someone else's Twin or from an Expert Twin you have access to.

Send to​

Who receives the output by email when the run completes. Type a name or email to add a recipient.

If you want a copy, add yourself to the list. Schedules deliver to the Send to recipients only — being the person who set up the schedule doesn't automatically put you on the list.

Email Subject (optional)​

The subject line on the email. Defaults to the Agent's name. Useful to customize when you have multiple schedules running off the same Agent (e.g., "Monday Standup Brief" vs "Friday Reflection").

Frequency​

Three options:

  • One-time — runs once at a specific date and time. Useful for a scheduled report tied to a deadline, or testing before committing to a recurrence.
  • Daily — runs every day at the chosen time.
  • Weekly — runs on the days you pick at the chosen time.

Time and Time Zone​

The time the schedule fires, in the time zone you select. The time zone defaults to yours; change it if the recipients are in a different zone and the timing matters.

Repeat on (Weekly only)​

For weekly schedules, pick which days. Multiple days are fine — a Monday-Wednesday-Friday recap, for example.

After You Save​

The schedule is active immediately. The next run fires at the configured time.

You can add multiple schedules on the same Agent. The same Document Summarizer can have a daily schedule for one team and a weekly schedule for another, each with different inputs and recipients.

To pause a schedule, edit it or delete it from the Schedules section.

Patterns That Work​

Weekly brag doc — every Friday at 5 PM, run an Agent that queries your Twin for the week's accomplishments and emails them to you. Use them to update a running doc you pull from at review time.

Monday morning brief — every Monday at 8 AM, run an Agent that pulls the week's calendar and open items, and sends a one-pager to you and your manager.

Customer account digest — every Wednesday at 9 AM, run an Agent against the account's Expert Twin and email the status to the account team.

End-of-day standup — daily at 6 PM, run an Agent that summarizes your day's work for the team's async standup channel. Pair with an MCP action if you want it posted to Slack directly.

Where the Output Goes​

The output lands in the inboxes of everyone on the Send to list.

If the Agent's prompt also invokes MCP actions — writing to a Google Doc, posting to Slack, creating a Jira ticket — those actions happen as part of the scheduled run too. The email is the delivery mechanism; the actions are part of the work.

For more on Agents in general, including how to set up the underlying Agent in the first place, see the Agents page.