Every reply shows its work.
OnRepeat drafts guest replies from your house manual, your calendar, and your past conversations. Every draft cites its sources. Review every reply, or let it send on its own. Your call, per listing.
Maria Alvarez
Seaside Cottage · Airbnb
Is there parking at the cottage? We land around noon, could we drop our bags early?
11:42 AM
Hi Maria! Yes, there is a private driveway that fits two cars.1 The unit is on track to be ready early tomorrow, so you can check in from 1pm.2 Arriving before that? Leave your bags on the covered porch.3
AI drafts can make mistakes. Review before sending.
Every marker shows its source: house facts, documents, and live checks against the calendar or a lock at draft time.
01
Tested before it talks to a guest
OnRepeat builds a simulated stay for each listing and asks the questions guests actually ask. Every answer is graded and its citations are checked one by one. You review anything below the bar, and your feedback becomes teachings the AI applies from then on.
02
You can see what it knows
Coverage per topic, verified counts, and an honest “Not set” where a policy is missing. Facts carry provenance: where they came from, and when.
03
Every review teaches it
In Review mode the inbox is a training loop, not just a queue.
04
Autopilot or human-in-the-loop
Three modes, per listing
Off, Review, or Auto. In Review, every draft waits for your approval; in Auto, replies send on their own. Run different listings in different modes, side by side.
Autopilot on a schedule
Auto overnight, Review during the day. Escalations and low-confidence drafts always wait for a human.
Configurable guardrails
Pricing, safety, and any topic you choose escalate to you every time. Sensitive capabilities like remote unlock stay off until you turn them on.
Nothing happens quietly
Every draft, approval, and send lands in the feed, timestamped. Review from the web app, the iOS app, or right in Slack.
Ready to automate your rentals?
See it run a real reservation, start to finish.