TravelGenie

Trip planner that coordinates flights, hotels, events, weather, currency, and traffic data through one request instead of ten browser tabs.

What it does

Give it a prompt like:

"Plan a trip to Banff, Alberta from Reston, Virginia, June 7-14, 2025. Budget $5000. We like hiking, museums, and dining."

Claude coordinates seven specialized servers to:

• Find flights and hotels within budget
• Pull weather forecasts for the trip dates
• Find local events matching the stated interests
• Convert currencies
• Check traffic and crowd patterns
• Build a day-by-day itinerary with a budget breakdown

Architecture

Each server is a standalone MCP server. Claude calls them in sequence, passing output from one into the next (geocode a location, then search flights/hotels near it, then check weather for those dates, etc.).

Server	Job
Flight	Search and compare flights
Hotel	Search accommodations
Event	Find local events and activities
Geocoder	Address ↔ coordinates
Weather	Forecasts, current conditions
Finance	Currency conversion, market data
Traffic & Crowd	Real-time location analysis

Setup

git clone https://github.com/AjaykumarBalakannan/TravelGenie.git
cd TravelGenie
pip install -r requirements.txt
cp .env.example .env
# add your API keys to .env

You'll need:

• SerpAPI — flights, hotels, events. Free tier at serpapi.com
• Anthropic — Claude API key from console.anthropic.com
• Slack (optional) — only needed for the Slack bot
• Open-Meteo and OpenStreetMap require no key

Running it

# Web UI
streamlit run streamlit_app.py

# Slack bot
python slack_bot.py

The web UI runs at http://localhost:8501.

Calling it directly

from claude_orchestrator import execute_tool

flights = execute_tool("search_flights",
    departure_id="IAD",
    arrival_id="YYC",
    outbound_date="2025-06-07",
    return_date="2025-06-14",
    adults=2
)

weather = execute_tool("get_weather_forecast",
    location="Banff, Alberta",
    forecast_days=7
)

Claude Desktop (MCP mode)

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "flight-search": {
      "command": "uv",
      "args": ["--directory", "/path/to/TravelGenie/servers/flight_server",
               "run", "python", "flight_server.py"],
      "env": {"SERPAPI_KEY": "your_key"}
    }
  }
}

Config location — macOS: ~/Library/Application Support/Claude/claude_desktop_config.json, Windows: %APPDATA%\Claude\claude_desktop_config.json.

Deployment

Docker:

docker build -t travelgenie .
docker run -d -p 8501:8501 --env-file .env travelgenie

AWS (ECS Fargate + DynamoDB + Secrets Manager + CloudWatch):

cd aws/cloudformation
aws cloudformation create-stack \
  --stack-name travelgenie \
  --template-body file://infrastructure.yaml \
  --parameters file://parameters.json \
  --capabilities CAPABILITY_NAMED_IAM

See aws/Run.md for the full setup. The mock_application/ directory has load-test tooling comparing EC2 vs. ECS Fargate under concurrent load — see documents/Travel_Genie_Report.pdf for the writeup and results.

Troubleshooting

• Servers not responding — check API keys in .env, confirm Python 3.12.1+
• Import errors — pip install -r requirements.txt --upgrade
• Slack bot not connecting — verify tokens, confirm Socket Mode is enabled, check bot has chat:write and app_mentions:read scopes
• No search results — try airport codes instead of city names; check SerpAPI quota; confirm dates are in the future

Testing

python slack_bot.py --test
streamlit run streamlit_app.py

License

MIT — see LICENSE.

Author

Ajaykumar Balakannan — github.com/AjaykumarBalakannan