An MCP (Model Context Protocol) server that exposes Skyscanner flight and airport search functionality to AI assistants like Claude Desktop and Cursor.
This project is experimental and intended for educational purposes only.
- This software is provided "as is" without any warranties
- Not intended for commercial use - Do not use this software in any commercial or production environment
- The Skyscanner API client library used here is reverse-engineered and may violate Skyscanner's Terms of Service
- Use at your own risk - The authors are not responsible for any consequences of using this software
- This project is for learning and experimentation with MCP servers and API integration
✈️ Flight Search - Search for flights between airports with flexible date options- 🛫 Airport Search - Find airports by name, city, or IATA code
- Python 3.8 or higher
- Git (for submodule support)
# Clone with submodules (recommended)
git clone --recursive https://github.com/shadyvb/mcp-skyscanner.git
cd mcp-skyscannerIf you already cloned without submodules:
git submodule update --init --recursivepip install -r requirements.txtWhat gets installed:
fastmcp- MCP server frameworkcurl_cffi,typeguard,orjson- Skyscanner API dependenciesskyscanner-api- Included as git submodule (automatically loaded bymcp_server.py)
The mcp_server.py automatically adds the vendor/skyscanner submodule to the Python path, so no additional installation steps are needed.
-
Locate your Claude Desktop configuration file:
- macOS:
~/Library/Application Support/Claude/claude_desktop_config.json - Windows:
%APPDATA%\Claude\claude_desktop_config.json - Linux:
~/.config/Claude/claude_desktop_config.json
- macOS:
-
Add the MCP server configuration:
Open the config file and add the following (or merge with existing
mcpServers):
{
"mcpServers": {
"skyscanner": {
"command": "python3",
"args": ["/absolute/path/to/mcp-skyscanner/mcp_server.py"],
"env": {
"SKYSCANNER_LOCALE": "en-US",
"SKYSCANNER_CURRENCY": "USD",
"SKYSCANNER_MARKET": "US"
}
}
}
}Important notes:
- Replace
/absolute/path/to/mcp-skyscanner/mcp_server.pywith the absolute path to yourmcp_server.pyfile - Use
python3instead ofpythonon macOS/Linux - If using a virtual environment, use the full path to Python (e.g.,
/path/to/venv/bin/python3)
-
Restart Claude Desktop completely (quit and reopen the application)
-
Verify installation: After restarting, you should see the Skyscanner tools available in Claude Desktop. Try asking: "Search for flights from LHR to JFK on December 22nd"
SKYSCANNER_LOCALE: Locale code (default: "en-US")SKYSCANNER_CURRENCY: Currency code (default: "USD")SKYSCANNER_MARKET: Market region code (default: "US")
Once configured, the following tools are available in Claude Desktop:
Search for airports by name, city, or IATA code.
Example usage in Claude:
- "Search for airports in London"
- "Find airports with code LHR"
- "What airports are in Cairo?"
Parameters:
query(string, required): Airport name, city, or IATA code (e.g., "London", "LHR", "Cairo")depart_date(string, optional): Departure date in ISO format (e.g., "2025-12-22")return_date(string, optional): Return date in ISO format
Returns: List of airport objects with title, skyId (IATA code), and entity_id
Search for flights between two airports.
Example usage in Claude:
- "Search for flights from London Heathrow to New York JFK on December 22nd"
- "Find flights from CAI to OPO on 2025-12-22"
- "Show me round-trip flights from LHR to JFK departing December 22, returning December 29"
Parameters:
origin_airport_code(string, required): IATA code (e.g., "LHR", "CAI", "JFK")destination_airport_code(string, required): IATA code (e.g., "JFK", "OPO", "LHR")depart_date(string, required): Departure date in ISO format (e.g., "2025-12-22" or "2025-12-22T10:00:00")return_date(string, optional): Return date for round-trip flightscabin_class(string, optional): "economy", "premium_economy", "business", or "first" (default: "economy")adults(integer, optional): Number of adult passengers, 1-8 (default: 1)
Returns: Dictionary with session_id and buckets containing flight options:
Best- Best overall optionsCheapest- Lowest price optionsFastest- Shortest duration optionsDirect- Non-stop flights (if available)
After setting up the MCP server in Claude Desktop, you can use natural language queries:
User: Search for flights from London Heathrow to New York JFK on December 22nd
Claude: I'll search for flights from London Heathrow (LHR) to New York JFK on December 22nd, 2025.
[Uses search_flights tool]
I found several flight options:
- Best: 8 options available
- Cheapest: 8 options available
- Fastest: 8 options available
- Direct: 3 non-stop options available
Would you like me to search for return flights as well?
- Check Python path: Use
python3instead ofpythonon macOS, or use full path to Python executable - Verify dependencies: Run
pip install -r requirements.txtto ensure all packages are installed - Initialize submodules: Run
git submodule update --init --recursiveif thevendor/skyscannerdirectory is empty - Check logs: Look for errors in
~/Library/Logs/Claude/mcp-server-skyscanner.log(macOS) - Test manually: Run
python3 /path/to/mcp_server.pyto check for import errors
spawn python ENOENT: Usepython3instead ofpythonin configModuleNotFoundError: No module named 'fastmcp': Runpip install -r requirements.txt- Date format errors: Ensure dates are in
YYYY-MM-DDformat and in the future
The server returns structured error responses:
BannedWithCaptcha: Rate limited by SkyscannerTimeout: Search exceeded maximum retriesInvalidDate: Date format or validation errorAirportNotFound: Airport code not found
- Flight searches may take 10-30 seconds (API polls for results)
- Dates must be in ISO 8601 format (
YYYY-MM-DDorYYYY-MM-DDTHH:MM:SS) - Dates must be in the future
- Airport codes are case-insensitive
GPL-3.0
Important: This license applies to the MCP server code only. The Skyscanner API client library (vendor/skyscanner) has its own license terms. Please review the license of the upstream repository before use.
- Skyscanner API client: irrisolto/skyscanner
This project is experimental and for educational purposes only. It is not intended for commercial use. Use at your own risk.