Multi-Agent Travel Planner

An autonomous multi-agent system that plans complete trips by coordinating specialized AI agents one searches flights, another searches hotels, and a third synthesizes everything into a personalized itinerary with booking links.

Architecture

• LangGraph orchestrates the pipeline as a state graph with parallel execution
• Flight Agent: queries Google Flights via SerpAPI
• Hotel Agent: queries Google Hotels via SerpAPI, runs concurrently with flights
• Itinerary Agent: GPT-4o-mini synthesizes results into a day-by-day plan
• Input validation: resolves free-text city names (e.g. "Karachi") to IATA airport codes automatically via LLM

Flight and hotel search run in parallel (fan-out/fan-in) rather than sequentially, cutting response time roughly in half compared to a naive sequential pipeline.

Tech Stack

Python · LangGraph · LangChain · OpenAI (GPT-4o-mini) · SerpAPI · FastAPI

Setup

git clone <repo>
cd travel-planner
python -m venv venv && source venv/bin/activate
pip install -r requirements.txt
cp .env.example .env  # add your API keys
uvicorn main:app --reload

API

POST /plan-trip

{
  "origin": "Karachi",
  "destination": "Dubai",
  "start_date": "2026-08-10",
  "end_date": "2026-08-17",
  "budget": 1500,
  "travelers": 1
}

Returns a full itinerary, ranked flight options, ranked hotel options, and any errors encountered.