Project

Wayfarer: AI travel planning agent

An AI-powered travel planning platform that turns preferences and destination ideas into personalised travel plans.

Try the live app

Overview

Planning a trip usually means moving between separate searches for flights, places to stay and things to do. Wayfarer brings those steps into one conversation. A traveller describes a destination or idea, and the agent helps shape it into a structured plan.

What it does

  • Turns user preferences and destination ideas into personalised travel plans.
  • Helps users explore flight options.
  • Helps users discover accommodation.
  • Generates structured travel itineraries.

The experience

Wayfarer is an interactive web application built to simplify end-to-end trip planning. It opens on a single question, "Where will you go next?", with one-click starting points such as a weekend in Rome or a beach escape in Bali. From there, the user chats with the planner in plain language, and replies appear in the same conversation.

The interface also reminds users that travel ideas are AI-generated and that flight details should be verified before booking.

Tech stack

  • LangGraph
  • LangChain
  • AI agents
  • Python
  • FastAPI
  • SQL
  • Flight and hotel APIs
  • Grok API

How the stack fits together

Wayfarer uses LangGraph and LangChain for its agent workflow, with Python and FastAPI on the backend and SQL for data. Flight and hotel APIs provide travel options, and the Grok API provides the language model. For more on this kind of system, read about AI orchestration with LangGraph.

How a travel agent like this works

This is the general pattern behind AI travel agents, rather than a line-by-line description of Wayfarer. A request arrives at an API, and an agent works out what the traveller wants. It then decides which tools to call, such as a flight search or a hotel search. The results go back to the language model, which turns them into a plan the traveller can read. A graph-based workflow keeps these steps in order and lets the agent ask a follow-up question when details such as dates are missing.

Because the options come from live travel data and the wording comes from a model, travellers should verify details before booking. Wayfarer's interface says the same.

Draft: details to confirm

Agent design

  • [Confirm] A planner agent reads the request and collects missing details such as dates and budget.
  • [Confirm] A flight agent and a hotel agent each call their travel API.
  • [Confirm] An itinerary agent combines the results into a day-by-day plan.
  • [Confirm] SQL stores conversations and saved trips.

Challenges

  • [Confirm] Keeping flight and hotel results in a consistent format for the model.
  • [Confirm] Handling vague requests by asking a follow-up question instead of guessing.

Deployment

The application runs on Render. You can try the live app.

Related services