Everything else on this page is request-response: you ask, it answers. The interesting travel problems are not shaped like that. A price moves three weeks before you fly. A tour you booked quietly gets cancelled. A host's strongest weekend suddenly has no bookings in it. Nobody is sitting there asking.
The designed answer is two agent loops on a real cadence rather than one more chat box. A trip-watch loop re-reads a traveller's trip plans and bookings on a schedule and raises only deltas: an item whose price or availability changed, a plan with a gap or a collision in it, a booking that needs an action before a date. A host-advisor loop does the mirror image over a provider's own listings and analytics, surfacing a utilization dip or a cancellation cluster while it is still early enough to act on.
The design constraints are the whole point. A loop that runs on its own must be quieter than one you invoke, so it reports differences rather than restating state, and it stays advisory — it raises a card, it does not rebook your holiday or reprice a room. It runs against the same governed graph with the same role scoping as the person it reports to, and its own runs are recorded, so "why did I get this?" always has an answer.
None of this exists yet, and the honest reason is that it depends on things that also do not exist yet: the planner and pricing engines that produce the judgements a loop would deliver, and the platform-level agent-run ledger that would record each scheduled execution and its cost. GGNomad does already have the pieces underneath — a scheduled-jobs surface, a notifications service, an append-only audit trail on listing and booking transitions, and analytics computed from live bookings.
Where this lands in the app is an AI operations surface: decision cards you accept or dismiss, an activity feed of what the agents did, and explicit confirmation gates in front of anything that writes. That surface is designed and ticketed too.
Illustrative scenario describing roadmap work. Nothing here runs today, and no prices, alerts, or trips are real. This story is narrative-only for now — there is no interactive "Do it yourself" walkthrough because the scheduled agent loops are still being built.
Ready to make this your story?
