My Travel Claw: From Prompt to Plan
I showed travelers how a personal AI assistant can remember their preferences and turn a short request into an emailed plan built around fixed commitments.
Community Meetup · Singapore · 28 April 2026
Planning a Trip
My example was a Hong Kong trip with conference days already reserved. I wanted inexpensive accommodation, a fast flight, and nearby attractions grouped together in the remaining time. The assistant used those priorities to shape the itinerary rather than offering the same destination list to every traveler.
At Agentic Night, I presented the journey from a request in Telegram to a tailored plan by email, and discussed alerts about sudden events affecting a trip. Booking and payment remained future work. The demonstration was about preparing a useful plan, not completing travel purchases.
Building an Itinerary Around Personal Priorities
Remembered Preferences and Reserved Days
The request covered 25 to 29 May in Hong Kong, with 26 to 28 May reserved for the conference. Those dates constrained the schedule, while remembered preferences guided choices such as inexpensive accommodation and a fast flight. Nearby attractions could be grouped into the remaining time instead of competing with the conference days.
From a Telegram Request to an Emailed Plan
Telegram supplied the request channel and email delivered the itinerary. The assistant brought the trip dates and personal priorities together into that planning output. The discussion also considered alerts about sudden events affecting a trip, extending the planning question from what to do to what may need reconsideration.
Booking and Payment as Future Work
Machine Payments Protocol, x402, and StableTravel with Shared Payment Tokens were discussed as possible future booking and payment paths. They were not implemented purchase capabilities in this demonstration. Delivering an itinerary therefore marked completion of the planning task, not confirmation of flights, accommodation, or payment.