Inspiration
We are out-of-state students who know the challenges of what should be a simple trip home. Flight prices are constantly fluctuating, getting a ride to the airport is a hassle, and classes can cause conflicts. We built Wing as a way to make this process as simple as texting a friend. Rather than juggling travel websites, ride apps, and your calendar, users can get their flight and ride to the airport all figured out within a single conversation.
What it does
Wing is a personal AI travel agent that you text over iMessage. Wing will ask the user for their home airport, budget, and to sync their calendar. Wing will then remember this information and find flights for you based on your preferences and schedule. Users can compare flight options and book a flight all within the conversation. After confirming a flight, Wing will add a reminder in your calendar and also give the option to book an Uber to the airport. The final trip itinerary, which includes trip status, arrival/ departure times and destinations, and Uber details, is presented in a static page.
How we built it
We used Photon's Spectrum as the iMessage infrastructure that lets us text Wing on our phones. Claude Sonnet 5.5 runs behind the scenes as the brain powering the agent. Using its respected API, Wing is able to connect to Google Calendar for a contextual understanding of the user's schedule. This lets it make suggestions about flights during weeks where a student is off from school, for example. We use Super Memory for continual learning, so Wing adapts to user's preferences and things it's been told from previous chats. OpenAI's coding agent Codex for used for most of the heavy implementation work.
Challenges we ran into
A challenge we ran into was figuring out how to demo booking the flight. While implementing payment is relatively trivial, it's unrealistic to test with purchasing actual flight tickets. Therefore we settled for a simulated payment structure that would mimic the process completed at a real checkout situation.
Accomplishments that we're proud of
We are proud of figuring out how Wing builds an understanding of the schedule given to it via Google Calendar. From this, it's able to suggest info like finding the cheapest flights to a vacation spot after your final exams end, for example.
What we learned
We learned how to ship an agent that runs natively on iMessage, which is quite different from most projects we have had previous experience with. Web apps and mobile apps require more of a focus on UI to ensure a good user experience, whereas a purely textual experience requires more attention to conversational details. We had to do a lot of narrowing to make Wing sound the way we wanted it to.
What's next for Wing
In the future, Wing will be expanded to work in group chats. Currently limited by the Pro plan on Photon, we cannot throw Wing into a group chat on Messages. With the Business plan, we would be able to let Wing coordinate travel plans with various people who want to take off from different airports. We view collaboration features as key to taking Wing's capabilities to the next level.
Built With
- claudeapi
- codex
- next.js
- photon
- supabase
- supermemory
- typescript
Log in or sign up for Devpost to join the conversation.