Webhooks give agents ears

The cost of hearing late

A reply from a hot lead sits in the inbox for six hours. A cancelled booking never triggers a rebooking attempt. A failed workflow waits for someone to notice on Monday. None of these are intelligence problems. They're hearing problems: nobody, human or agent, knew the moment it happened.

The naive fix is an agent that polls: check the inbox every minute, forever. Now you're paying an AI to discover, thousands of times a day, that nothing happened. The right fix is older and cheaper: events. The world tells you when something changes.

How an agent with ears works

In AgentsHQ, a webhook is owned by a specific agent and carries its own instructions. When our CRM receives an email reply, it notifies the follow-up specialist agent. When a booking is cancelled, the scheduler agent hears about it. Each event wakes exactly one agent, briefed for exactly that situation.

The parts that make it safe enough for a business:

• Signed and unguessable. Every webhook lives at its own random URL and every delivery is verified with a cryptographic signature over the raw payload. A message either proves who sent it or it's rejected.
• A briefing, not raw data. The event arrives inside instructions for that agent: what this event means, what to do, and what never to do. Ours end with the rule that matters: never send email; draft, and let a human review.
• A fresh session per event. Event number 147 can't inherit confusion from event number 12. Every wake-up starts clean.
• No double-fires. Senders retry. Deliveries are recorded under a uniqueness guarantee, so a retried event can't trigger the same action twice. Boring, and exactly what keeps an agent from double-replying to a customer.

What changes for your team

Minutes-fast first responses, drafted and waiting in Review before you'd have seen the notification. Rebooking attempts that start themselves. Failures that page an agent instead of waiting for a human to browse a dashboard. And through all of it, spend that scales with events that actually happened, not with hours on a clock.

Ears, then a calendar, then hands: webhooks for the world's timing, cron for yours, and the task board in the middle so everything an agent does lands where a human can see it, review it, and trust it.

---

We run our own business on AgentsHQ: a fleet of AI agents with written instructions, a shared task board, schedules with circuit breakers, and a human review gate before anything ships. If you want agents doing real work for your team, without giving up control, start at agentshq.dev.