Triggers and schedules

Not every agent should be run by hand. RunAIAgents gives you several ways to start a run automatically, so an agent reacts to events or runs on a schedule without you opening the app.

Webhooks

An inbound webhook lets an external system trigger a run by sending an HTTP request. Every webhook is HMAC-verified — the sender signs the payload with a shared secret, and RunAIAgents checks that signature before starting a run, so an attacker who finds the URL can't trigger runs without the secret.

Use a webhook when the source system can push events to you as they happen — a CI pipeline finishing, a form submission from a third-party tool, a CRM's own webhook feature.

Authenticated polling

Some services don't support outbound webhooks, or you don't control the sending side. Authenticated polling has RunAIAgents check the service on an interval, using your connected credentials, and start a run when there's new data to act on.

Use polling when the source system can't push to you but exposes an API you can read from — a legacy system, or a service where webhooks aren't available on your plan.

Cron schedules

A schedule starts a run on a recurring basis using a standard cron expression — every hour, every weekday morning, the first of the month. Use a schedule for work that doesn't depend on an external event at all: a daily summary, a weekly report, a periodic cleanup task.

Channels

Channels let people and other systems start a run by sending a message, rather than by calling an API. Supported channels are:

• Inbound email — a run starts when an email arrives at an address tied to the agent.
• Telegram
• WhatsApp
• Web forms
• An embeddable widget you can drop into your own site or app.

Use a channel when the trigger is a person interacting directly — asking a question, submitting a request — rather than a system event.

Choosing between them

As a rule of thumb:

• Reacting to an event from a system you don't control that supports webhooks → webhook.
• Reacting to an event from a system that doesn't support webhooks → polling.
• Running on a fixed cadence regardless of events → cron schedule.
• Letting a person start the agent directly → a channel.

An agent isn't limited to one trigger — you can wire a schedule for a recurring check and a webhook for real-time events on the same agent.

Next steps

Once a trigger fires and a run starts, see running agents for how a run actually executes and how to watch it happen.