BuddyTravelAI
At Bravent, we have developed BuddyTravelAI, an intelligent travel recommender built with AG-UI and Microsoft Agent Framework.
BuddyTravelAI is an agentic application, meaning that AI is not just a chat interface, but a true intelligent assistant capable of collaborating with you throughout the entire travel planning process.
This approach represents the evolution of digital interfaces: moving from traditional chatbots to Adaptive Agentic Interfaces, the new way of interacting with applications, where the interface dynamically adapts to the agent’s reasoning and state.
BuddyTravelAI:
-
Understands what you need and builds step-by-step itineraries.
-
Searches for and adjusts accommodations based on real user criteria.
-
Adapts to your travel style, remembering budget, duration, trip type, and personal preferences.
-
Automatically updates the interface as the agent progresses.
-
Refines recommendations when filters or preferences change.
-
Asks for missing information to deliver a more accurate plan.
-
Allows you to fine-tune the trip through chat, as if speaking with an expert.
-
Generates an itinerary ready to print, save, or share.
Below, we explain how BuddyTravelAI is built and how it works, from the perspective of an Adaptive Agentic Interface.
Core Components of BuddyTravelAI
BuddyTravelAI is composed of the following key components:
1. Frontend (UI)
The frontend is the visible part of the application: buttons, lists, destinations, and progress messages.
Its role goes beyond displaying information—it reflects in real time every decision made by the agent.
Thanks to AG-UI, the interface is not static: it updates automatically based on agent-generated events, showing progress, intermediate states, and new recommendations as they are produced.
2. Backend: Microsoft Agent Framework
The backend is the brain of the application.
It receives user requests, processes them, and returns structured responses to the interface.
This is where the agent lives, responsible for reasoning, remembering preferences, maintaining context, and generating travel recommendations.
The backend is built in .NET using Microsoft Agent Framework (MAF), Microsoft’s framework for building advanced AI agents.
What Does Microsoft Agent Framework Do Inside the Application?
1. Step-by-step reasoning: MAF allows the agent to process a traveler’s request as a structured reasoning flow, not as an isolated message.
The agent analyzes the destination, trip type, and duration, compares options, and generates personalized itineraries, adjusting results according to user preferences.
2. AI model usage: MAF enables the agent to connect to Azure OpenAI to process user requests and generate intelligent responses.
3. Agent memory and state: A defining characteristic of Adaptive Agentic Interfaces is memory.
The agent can retain and recall relevant information throughout the entire planning process thanks to agent sessions and an integrated memory system provided by MAF.
These sessions store state and context, including preferences, applied filters, chat interactions, and itinerary progress—allowing the agent to reason coherently over time without losing information.
4. Emitting AG-UI events to update the interface: Communication between frontend and backend is based on an event-driven AG-UI protocol.
This keeps the agent and the interface fully synchronized in real time.
When the agent updates something—such as “searching destinations…”, “analyzing weather…”, or generating new options—the UI reflects it instantly.
MAF enables the agent to emit these events, making a live, adaptive interface possible.
How BuddyTravelAI Works
BuddyTravelAI is designed as an intelligent co-pilot that guides users through travel planning step by step.
1. Enter the trip destination: At the top of the application, you enter the trip title, such as “Trip to Bali – Spring 2026” or “Paris Getaway”.
BuddyTravelAI automatically detects the destination from the title.
2. Add basic trip details: Next, you define two essential elements:
-
Trip duration (number of days)
-
Trip type (romantic, adventure, family, etc.)
3. Submit the request: When you click the plan button:
-
The agent analyzes the destination, duration, and travel style
-
Generates a personalized day-by-day itinerary
-
Provides general travel recommendations such as activities, tips, and key places
4. Add accommodation preferences: In the accommodation preferences section, you can specify:
-
Accommodation type
-
Price range
-
Minimum rating
-
Location
-
Board type (breakfast, half board, etc.)
-
Number of travelers
-
Services (kitchen, private bathroom, free cancellation, etc.)
BuddyTravelAI analyzes your profile, automatically updates recommendations, and suggests hotels aligned with your criteria.
5. Make the plan truly personal: Beyond essential travel and accommodation data, you can add additional requirements to further refine the plan.
These requirements may relate to activities, transportation, accommodation, or personal preferences (for example: “Add an extra day for rest”).
Each new requirement triggers a real-time update of the itinerary and recommendations, adapting the trip precisely to your needs.
6. The right-side chat: your travel co-pilot: On the right side of the interface, an integrated chat allows you to continue refining your trip.
Simply type what you need—
“Show cheaper options”, “Add activities for kids”, “Closer to the city center”, “Find a hotel with a spa”—
and BuddyTravelAI adapts the plan accordingly.
This chat is not a chatbot, but a continuous collaboration channel with the agent.
7. A tailor-made plan, ready to print: Finally, the application allows you to export your itinerary as a PDF document, ready to save, carry with you, or easily share with your travel companions.
Conclusion
BuddyTravelAI is not just a travel application.
It is a concrete example of how Adaptive Agentic Interfaces are redefining the way we interact with software.
This is not about talking to AI.
It is about working with it.
