TripMaxxing

Inspiration

Planning travel is often overwhelming—endless tabs, fragmented tools, and decision fatigue. We wanted to create a single, intelligent assistant that makes trip planning feel effortless, enjoyable, and personal. With recent breakthroughs in AI agents and conversational interfaces, we saw an opportunity to transform how people plan trips—by making it feel more like chatting with a well-informed local expert.

What it does

TripMaxxing is a personalized itinerary tailoring assistant that helps you:

🧠 Chat with an AI agent to plan your trip conversationally
🧭 Take a travel persona quiz to define your preferences
📅 View, organize, and manage your trip itinerary day-by-day
✈️ Get suggestions for flights, hotels, food, activities, and local gems
💸 Track your travel budget and receive cost-aware suggestions

Behind the scenes, TripMaxxing uses a multi-agent AI backend (CrewAI) with Gemini 2.5 Flash to orchestrate a collaborative planning team: a Trip Planner, Booking Agent, and Local Expert. Exa – an AI-powered search engine to discover relevant events, activities, and suggestions across the web. Browserbase – a headless browser to scrape dynamic content from sites like Pinterest, Airbnb, and Eventbrite, ensuring we always have the freshest, most visually rich data.

How we built it

Frontend:

Next.js 15 (App Router)
Tailwind CSS for styling
TypeScript + React Hook Form + Zod for robust user flows
Lucide Icons for clean visuals
Mobile-first, responsive design

Backend:

Python 3.12
CrewAI for managing AI agent workflows
Google Gemini 2.5 Flash LLM for fast, smart responses
Custom travel tools for destination search, booking suggestions, and activity recommendations
W&B for logging the model path and agent calls

Challenges we ran into

Agent Orchestration: Designing modular, cooperative agents that don’t overlap or contradict each other was tricky.
Response Time: Making LLM-based agents fast enough for real-time UX without degrading planning quality took a lot of tuning.
User Preferences → Itinerary: Translating quiz answers and chat context into coherent, day-by-day itineraries required building a shared memory system for agents.
Frontend Complexity: Integrating dynamic forms, chat UX, and trip state management cleanly in Next.js was a constant juggling act.

Accomplishments that we're proud of

Built a fully functional multi-agent travel planning system powered by Gemini in just one weekend
Created a delightful, responsive mobile-first interface with beautiful UX
Designed an intuitive travel quiz that generates useful, persona-based recommendations
Seamlessly integrated CrewAI agents into a cohesive travel planning workflow

What we learned

LLMs + structured tooling can deliver surprisingly reliable planning workflows when modularized correctly.
Conversational planning is a powerful UX metaphor—it lowers barriers and builds user trust.
Good design matters. A clean UI radically enhances the perception and usability of AI tools.
CrewAI is a powerful framework but needs careful orchestration logic to avoid redundant or conflicting agent outputs.