How agents run: manual, scheduled & triggered Guide
The three ways an agent can start — you run it, it runs on a schedule, or a connected app triggers it automatically.
Last updated July 16, 2026
On this page
An agent can start in three ways: you run it yourself, it runs on a schedule you set, or a connected app starts it automatically when something happens. This article explains each option and how to set them up.
Run an agent yourself
The simplest way to start an agent is to run it on demand. Open the agent and start a run whenever you want it to work. This is ideal while you're building and testing, or for tasks you only need occasionally.
Running an agent yourself gives you full control over timing and lets you confirm it behaves the way you expect before you automate it.
Tip: Test a new agent with a manual run or two before putting it on a schedule or connecting a trigger. Once you're confident in the results, automate it.
Run an agent on a schedule
When a task needs to happen regularly — a daily summary, a weekly report, a recurring check — you can have the agent run on a schedule instead of starting it by hand.
Scheduled runs are best for predictable, time-based work: anything you'd otherwise remember to trigger yourself at the same time each day or week. The agent runs on its own and delivers its results without you needing to be present.
Let a connected app trigger an agent
The most powerful option is an automatic trigger. When you connect one of your own tools from the Integrations hub, some apps can start an agent the moment a specific event happens — no scheduling, no manual step. The agent responds in near real time to what's happening in your business.
Examples of triggers
Which events are available depends on the app. Common examples include:
- A new order is placed in your e-commerce store.
- A record is created or updated in your CRM.
- A bank or payments event occurs.
- An inbound text message or phone call arrives on a connected number.
Not every app offers triggers. Many apps only provide actions — things the assistant and your agents can do — without any events that start an agent on their own. Triggers are specifically the events that make an agent run by itself.
Check whether an app can trigger an agent
Before you rely on an app to start an agent, confirm it offers the trigger you need:
- Open the Integrations hub and find the app you're interested in.
- Open its capabilities view to see its Actions and, where available, its Triggers.
- Search the list for the event you have in mind to confirm it's supported.
If an app lists no triggers, it can still be used by your agents for actions — you'll just start those agents manually or on a schedule instead.
Set up an automatic trigger
To have a connected app start an agent automatically, work through these steps:
- Connect the app. From the Integrations hub, connect the tool you want and complete its sign-in or setup flow.
- Enable the actions the agent needs. When connecting, choose which of the app's actions the assistant and your agents are allowed to perform. You can adjust these later from the connection's Manage actions view.
- Confirm the trigger exists. Check the app's capabilities view to make sure the event you want is available as a trigger.
- Configure the agent to start on that event. Set the agent up to run when the trigger fires, so it responds automatically each time the event happens.
Note: Triggers only work for apps you've connected in your own workspace, and only for events that app supports. If an agent isn't starting automatically as expected, confirm the app is still connected and that the trigger you're relying on is listed in its capabilities.
Choosing the right approach
| Start method | Best for |
|---|---|
| Run it yourself | Testing, one-off tasks, and work you prefer to start manually |
| On a schedule | Predictable, recurring work at a set time — daily, weekly, or otherwise |
| Connected-app trigger | Reacting instantly to events in your tools, like new orders or inbound messages |
Many teams combine all three: they test an agent with a manual run, put routine work on a schedule, and connect triggers so the agent can respond the moment something happens.