Triggers and schedules

Agents do not have to wait for you. There are two ways work starts on its own: at a time you choose (a schedule), or when something happens in a connected account (a trigger).

Both are set up per agent, in Agent Settings, and need the Editor or Admin role. Every automated run produces an ordinary chat, so you can read back what happened.

Schedules

A schedule runs an agent at a set time. In Agent Settings, open Schedules and choose Add schedule:

• Frequency: One time, Daily, Weekly or Monthly.
• When: for a one-time run, the date and time. For the others, the time of day, plus the days of the week (weekly) or the day of the month (monthly).
• First message: the message that starts each run, which is what the agent is asked to do, for example "Email me today's meetings with attendees and Meet links."
• Enabled: switch a schedule off to pause it without deleting it.

Times are in your computer's time zone, which is shown next to the time and kept when you edit the schedule later. A monthly schedule skips the months that do not have its day, so a schedule for the 31st does not run in April. A one-time schedule switches itself off after it has run.

You can also create, change and delete schedules by describing them in the agent's admin chat.

Triggers

A trigger runs an agent when something happens in a connected account. There are two: a new email in Gmail, and a new file in a Google Drive folder. In Agent Settings, open Integrations, open the settings of the account, go to the Triggers tab and choose Add trigger. Give it a Name, pick its Type and fill in its filters, so it fires only on what you care about. Its Enabled switch pauses a trigger without deleting it.

For example, a Gmail trigger can be limited by sender, subject, text in the body and whether the email has an attachment. Sender and subject are patterns, where * stands for any characters: a sender of *@acme.com matches everyone at acme.com, and a subject of *invoice* matches every subject that contains the word. Each connector's page lists its triggers and their filters; see the connector reference.

A trigger belongs to one agent and one connected account. A trigger is only available if the account granted the permission it needs; if one shows as unavailable, reconnect the account and allow the missing permission.

What happens when one fires

The agent starts a new chat and works through the task on its own, using the tools you allowed it. Chats started by a trigger show the connector's icon in the agent's chat list.

• The same event never starts two runs. One email cannot start a trigger twice.
• Tools that require approval still require it. The run pauses and waits for you instead of acting without permission. Tunori does not notify you, so check the agent's chats and answer there.
• The agent does not ask questions. Nobody is there to answer, so it works with what the trigger or schedule gave it; put anything it would otherwise ask about into its instructions.
• Runs are skipped when your credit runs out. They are not queued: a run that was skipped does not happen later. See billing and credits.
• Triggers need a working connection. If an account is disconnected or needs signing in again, its triggers stop until you reconnect it.

Choosing between them

Use a schedule when the timing is yours: a Monday briefing, a month-end report. Use a trigger when the timing belongs to someone else: a customer emailing you, a document being dropped in a folder.