What it does

A Python library and MCP server that searches Google Flights via reverse-engineered API calls rather than web scraping. Exposes two main tools: search_flights for finding flights on a specific date with detailed filters (origin, destination, cabin class, stops, airlines, layover duration), and search_dates for scanning a date range to identify the cheapest travel windows. Results include fare, flight duration, number of stops, and departure and arrival times. Direct API access avoids the brittleness of HTML parsing and browser automation.

Who it's for

Travel automation engineers building flight-search features into apps; backend teams wiring Claude into custom booking systems; anyone building AI-driven travel planners that need programmatic access to real-time Google Flights pricing without the maintenance burden of web scraping.

Common use cases

Find round-trip or one-way flights between two cities on a specific date, filtered by cabin class and connection preferences.
Scan a flexible date range to identify the cheapest travel dates—for example, finding the best week to fly from LAX to JFK in July.
Restrict results to specific airlines or global alliances, or exclude budget carriers from the results.
Compare flights across multiple currencies and languages by setting the currency code and language locale.
Find connecting flights with custom layover constraints (e.g., minimum 2 hours, maximum 4 hours in a hub).

Setup pitfalls

One secret was detected in the source code—audit and rotate any exposed credentials before use in production.
CI is currently failing; the project is actively maintained (last update 15 days ago), but review recent issues or commit messages to confirm your use case is not blocked.
Requires network access to Google Flights endpoints; proxy or firewall rules may need adjustment to allow outbound connections.
MCP server runs as fli-mcp (stdio) or fli-mcp-http (HTTP on port 8000)—verify the correct binary is in your PATH before connecting to Claude Desktop.