Inspiration

The idea for a2b came when my friends and I tried planning a group trip to Vietnam. We all had different budgets, schedules, and trip goals—some wanted a street food adventure, others were excited about the nightlife, and a few just wanted to relax. What followed was endless, frustrating discussions that eventually led nowhere because no one took initiative. That’s when the idea for an agentic travel planner was born—something that doesn’t just suggest plans, but actually takes action to make them happen.

What it does

a2b is an agentic travel planner that talks to each group member individually to understand their preferences—likes, dislikes, budgets, and interests. It then builds a personalized trip plan that tries to satisfy everyone. Once the itinerary is created, all group members vote on it. If the vote doesn’t pass, A2B regenerates the itinerary based on feedback. It doesn’t stop at planning—it books flights and hotels, only asking the user for payment.

How we built it

** is built as a full-stack Next.js application. It uses:

  • ElevenLabs for natural, conversational AI to talk to users
  • Gemini for generative reasoning and itinerary creation
  • Playwright for flight and hotel booking automation
  • Supabase to store user data and preferences
  • Netlify for hosting
    We started building on June 12th using the Bolt framework and have been actively developing it since.

Challenges we ran into

  • Integrating ElevenLabs conversational flow and making it sound natural
  • Ensuring flight booking automation was both fast and reliable
  • Hosting constraints: the project grew too large for single-instance hosting, so we had to modularize and host parts separately

Accomplishments that we're proud of

We’re proud that we would actually use A2B ourselves. It’s not just another itinerary planner—it’s an agentic one. It takes real actions to get a trip out of the group chat and into reality.

What we learned

AI isn’t just for chatting—it’s a powerful tool for automation and execution. We learned how to combine conversational AI with real-world action, bridging the gap between planning and doing.

What’s next for A2B

We’re working on supporting more versatile and dynamic trip types. While we currently support round-trip flights and multi-layovers, our goal is to enable real-time, adaptive trip planning that works on the go.

Built With

  • bolt
  • elevenlabs
  • gemini
  • nextjs
  • playwright
  • supabase
Share this project:

Updates