Custom triggers created with AI

Built-in triggers watch one app for one kind of event. Sometimes you need more: "tell me when a customer with an open support ticket also opens a new sales deal", or an app with no trigger at all. Create With AI triggers handle this.

You describe what to watch for, in plain words. The agent then:

• 1. connects and checks the services needed,
• 2. finds the exact tools each service offers,
• 3. writes a small program that polls (checks) those services for your condition,
• 4. tests it in an isolated sandbox and reports problems,
• 5. on its first run records a baseline (what already exists) without firing, so old items don't set it off,
• 6. after activation, checks on schedule and fires only for new matching items.

These triggers only read; they can detect conditions but never change your data. Any action happens afterwards, done by the agent.

Cost is the key difference from built-in triggers. Every check runs the trigger's code and uses credits, whether or not anything is found. Checks can run from every 5 minutes up to once a week. The agent shows an estimated polling cost before you activate. To save credits: check less often, keep the condition simple, and use a built-in trigger when one exists.

Guards: a limit on active custom triggers per agent, a circuit breaker that switches off a trigger that fires too many times in a short period, and polling pauses if credits run out.