gem-hunter

A powerful Python library that provides programmatic access to Google Flights data with an elegant CLI interface. Search flights, find the best deals, and filter results with ease.

🚀 What makes fli special? Unlike other flight search libraries that rely on web scraping, Fli directly interacts with Google Flights' API through reverse engineering. This means:

• Fast: Direct API access means faster, more reliable results
• Zero Scraping: No HTML parsing, no browser automation, just pure API interaction
• Reliable: Less prone to breaking from UI changes
• Modular: Extensible architecture for easy customization and integration

pipx install flights

# Run the MCP server on STDIO
fli-mcp

# Run the MCP server over HTTP (streamable)
fli-mcp-http  # serves at http://127.0.0.1:8000/mcp/

{
  "mcpServers": {
    "fli": {
      "command": "/Users/<user>/.local/bin/fli-mcp"
    }
  }
}

Note: Replace <user> with your actual username. You can also find the path to the MCP server by running which fli-mcp in your terminal.

The MCP server provides two main tools:

Total travelers (passengers + children + infants_in_seat + infants_on_lap) must be between 1 and 9, and infants_on_lap cannot exceed passengers.

A round trip costs 1 + top_n page fetches (one outbound search, plus one per outbound candidate expanded into return flights), so top_n is capped at 10. Round-trip results all from one airline? Raise top_n, or sort differently — the default sort only ever expands the cheapest top_n outbounds, which are often the same carrier.

Same passenger limits as search_flights: total 1-9, infants_on_lap ≤ passengers.

pip install flights

# Install using pipx (recommended for CLI)
pipx install flights

# Get started with CLI
fli --help

• 🔍 Powerful Search

  • One-way flight searches
  • Multi-city flight searches
  • Flexible departure times
  • Multi-airline support
  • Cabin class selection
  • Stop preferences
  • Custom result sorting

• 💺 Cabin Classes

  • Economy
  • Premium Economy
  • Business
  • First

• 🎯 Smart Sorting

  • Price
  • Duration
  • Departure Time
  • Arrival Time

• 🛡️ Built-in Protection

  • Rate limiting
  • Automatic retries
  • Comprehensive error handling
  • Input validation

Searches are served by Google's public search page rather than the FlightsFrontendService RPC. Since 2026-08 GetShoppingResults and GetCalendarGraph require an x-goog-batchexecute-bgr header that only the page's own JavaScript can produce, so a plain HTTP client gets HTTP 200 with no payload. Fli issues GET https://www.google.com/travel/flights?tfs=<protobuf> instead and reads the results out of the page's inline AF_initDataCallback blob keyed ds:1.

What that means in practice:

• Three filters are not supported. emissions, bags and exclude_basic_economy have no tfs field and cannot be reconstructed from the decoded rows, so they are dropped with a warning. Stops, cabin, passengers, alliances and layover bounds ride in the request; airline include/exclude, price cap, max duration and departure windows are applied to the results after fetching.
• Multi-city raises SearchUnsupportedError. Google loads those results client-side through the gated RPC, so the page carries no rows to read. Search each leg separately.
• get_booking_options is unavailable. It calls GetBookingResults, which is gated the same way, and currently raises SearchRejectedError. The per-flight tfs booking deep links are built offline and still work.
• Fewer rows per search. Expect roughly 20-45 itineraries, fewer than the old RPC returned — and a client-side filter cannot back-fill the list the way Google's server-side one did.
• Children and infants thin the results — sometimes to nothing. Google prices those parties client-side, so the page inlines fewer itineraries for them, and in premium cabins often none (measured 2026-09: JFK→LHR economy 23 rows for one adult, 16 with an infant; SFO→NRT business 9 rows for one or two adults, 0 with a child). Extra adults cost nothing. An empty result for such a search does not mean the route has no flights — and neither does it mean this is why: search() only warns when the fetched page itself came back with zero rows, not when the caller's own airline/price/ duration/window filter removed rows Google did return. An adults-only search shows the schedule.
• Date searches cost one page fetch per date. The page has no calendar grid, so a range is priced date by date; one SearchDates.search covers at most 93 dates and a wider range raises ValueError. Budget for it: 93 dates across 10 workers is several hundred MB of pages and parsed JSON at peak. A sweep that never manages to load a single page — the shape a blocked or consent-gated client produces — gives up after a handful of dates rather than paying the retry budget on all of them. Only pages served without results count towards that: a timeout or a dropped connection says nothing about the dates not yet tried, so those never abandon a sweep. Measured with the real backoff: 42 page fetches (bounded at 45, so up to ~135 HTTP requests once the client's own retries multiply in) and about 4 seconds, the same whether the range is 30 days or 93. Unbroken, a 93-date range would have cost 279 fetches and up to 837 requests. The bound is (5 + worker count) x 3, so raising configure_concurrency raises it proportionally. That breaker disarms for good the moment any page loads, even an empty one, so it cannot catch a sweep that is mostly timeouts around one lucky date — SearchDates.search raises that case too, whenever nothing priced and at least half the attempted dates never loaded. A minority of failures alongside real results, or alongside a confirmed-empty range (None), still returns normally but logs one warning naming the counts.
• A page occasionally arrives without results. Roughly one request in sixty returns HTTP 200 with no ds:1 blob; the client retries that case up to twice (0.5s then 1.5s) before raising SearchParseError. A healthy search never pays for it.
• FLI_SOCS_COOKIE. EU/EEA IPs are redirected to Google's consent interstitial, which serves no ds:1 blob. The client sends a pre-accepted SOCS consent cookie by default; set FLI_SOCS_COOKIE to change the value, or to an empty string to send none.
• FLI_CA_BUNDLE / CURL_CA_BUNDLE / REQUESTS_CA_BUNDLE. Behind a TLS-intercepting corporate proxy, set one of these (checked in that order) to a PEM CA bundle path. A bad or missing path, or a certificate the bundle doesn't cover, raises SearchCertificateError naming the variable — the client does not retry it, since a fixed bundle either verifies or it doesn't. Read once per worker thread on its first request; changing the value doesn't affect a thread's already-created session.

# Basic flight search
fli flights JFK LHR 2026-10-25

# Advanced search with filters
fli flights JFK LHR 2026-10-25 \
    --time 6-20 \             # Departure time window (6 AM - 8 PM)
    --airlines BA,KL \        # Airlines (British Airways, KLM)
    --class BUSINESS \        # Cabin class
    --stops NON_STOP \        # Non-stop flights only
    --sort DURATION           # Sort by duration

# Alliance + exclude + locale (May-2026 filter additions)
fli flights JFK LHR 2026-10-25 \
    --alliance ONEWORLD \
    --exclude-airlines AA \
    --min-layover 90 \
    --max-layover 360 \
    --currency EUR --language en-GB --country GB

# Round trip: raise --top-n to see more airlines (default 5, max 10, costs 1 + top_n
# page fetches per search)
fli flights JFK LHR 2026-10-25 --return 2026-11-01 --top-n 8

# Family mix: 2 adults, 1 child, 1 lap infant (total must be 1-9)
fli flights JFK LHR 2026-10-25 --passengers 2 --children 1 --infants-on-lap 1

⚠️ Experimental --format json is experimental. The JSON schema may change while the machine-readable CLI contract settles.

# Return machine-readable flight results
fli flights JFK LHR 2026-10-25 --format json

# Basic date search
fli dates JFK LHR

# Advanced search with date range
fli dates JFK LHR \
    --from 2026-01-01 \
    --to 2026-02-01 \
    --monday --friday      # Only Mondays and Fridays

⚠️ Experimental --format json is experimental for date searches as well.

# Return machine-readable date search results
fli dates JFK LHR --from 2026-01-01 --to 2026-02-01 --format json

# Two-leg multi-city trip
fli multi --leg SEA,HKG,2026-12-26 --leg PEK,SEA,2027-01-02

# Three-leg multi-city trip with filters
fli multi \
    -l SEA,NRT,2026-12-26 \
    -l NRT,HKG,2026-12-30 \
    -l HKG,SEA,2027-01-05 \
    --class BUSINESS \
    --stops 0

Total passengers (adults + children + infants) must be between 1 and 9.

--top-n only applies to round trips (it is rejected if set on a one-way search) and costs 1 + top_n page fetches. Round-trip results all from one airline? Raise --top-n (up to 10), or --sort differently — the default sort only expands the cheapest top_n outbounds, which are often the same carrier.

Fli includes a Model Context Protocol (MCP) server that allows AI assistants like Claude to search for flights directly. This enables natural language flight search through conversation.

# Run the MCP server on STDIO
fli-mcp

# Or with uv (for development)
uv run fli-mcp

# Or with make (for development)
make mcp

# Run the MCP server over HTTP (streamable)
fli-mcp-http  # serves at http://127.0.0.1:8000/mcp/

To use the flight search capabilities in Claude Desktop, add this configuration to your claude_desktop_config.json:

Location: ~/Library/Application Support/Claude/claude_desktop_config.json (macOS)

{
  "mcpServers": {
    "flight-search": {
      "command": "fli-mcp",
      "args": []
    }
  }
}

After adding this configuration:

1. Restart Claude Desktop
2. You can now ask Claude to search for flights naturally:
   • "Find flights from JFK to LAX on December 25th"
   • "What are the cheapest dates to fly from NYC to London in January?"
   • "Search for business class flights from SFO to NRT with no stops"

from datetime import datetime, timedelta
from fli.models import (
    Airport,
    PassengerInfo,
    SeatType,
    MaxStops,
    SortBy,
    FlightSearchFilters,
    FlightSegment
)
from fli.search import SearchFlights

# Create search filters
filters = FlightSearchFilters(
    passenger_info=PassengerInfo(adults=1),
    flight_segments=[
        FlightSegment(
            departure_airport=[[Airport.JFK, 0]],
            arrival_airport=[[Airport.LAX, 0]],
            travel_date=(datetime.now() + timedelta(days=30)).strftime("%Y-%m-%d"),
        )
    ],
    seat_type=SeatType.ECONOMY,
    stops=MaxStops.NON_STOP,
    sort_by=SortBy.CHEAPEST,
)

# Search flights
search = SearchFlights()
flights = search.search(filters)

# Process results
for flight in flights:
    print(f"💰 Price: ${flight.price}")
    print(f"⏱️ Duration: {flight.duration} minutes")
    print(f"✈️ Stops: {flight.stops}")

    for leg in flight.legs:
        print(f"\n🛫 Flight: {leg.airline.value} {leg.flight_number}")
        print(f"📍 From: {leg.departure_airport.value} at {leg.departure_datetime}")
        print(f"📍 To: {leg.arrival_airport.value} at {leg.arrival_datetime}")

Runnable examples live in examples/python/:

# Run examples with uv (recommended)
uv run python examples/python/basic_one_way_search.py
uv run python examples/python/round_trip_search.py
uv run python examples/python/date_range_search.py

# Or install dependencies first, then run directly
pip install pydantic curl_cffi httpx
python examples/python/basic_one_way_search.py

Available Examples:

• basic_one_way_search.py - Simple one-way flight search
• round_trip_search.py - Round-trip flight booking
• date_range_search.py - Find cheapest dates
• multi_city_search.py - Multi-city itinerary across several legs
• advanced_filters_search.py - Alliances, airline exclusions, layovers, locale
• complex_flight_search.py - Advanced filtering and multi-passenger
• time_restrictions_search.py - Time-based filtering
• date_search_with_preferences.py - Weekend filtering
• price_tracking.py - Price monitoring over time
• error_handling_with_retries.py - Robust error handling
• result_processing.py - Data analysis with pandas
• complex_round_trip_validation.py - Advanced round-trip with validation
• advanced_date_search_validation.py - Complex date search with filtering

Examples are organized by language, with parallel scripts so you can compare the two APIs:

• Python — examples/python/
• TypeScript — examples/typescript/

# Python
uv run python examples/python/complex_flight_search.py

# TypeScript (from examples/typescript, after `bun install`)
bun run multi_city_search.ts

Example Categories:

• Basic Usage: One-way, round-trip, date searches
• Advanced Filtering: Time restrictions, airlines, alliances, seat classes, locale
• Data Analysis: Price tracking, result processing with pandas
• Error Handling: Retry logic, robust error management
• Complex Scenarios: Multi-city, multi-passenger, validation, business rules

Each example is self-contained — change the airports, dates, and filters at the top of the script to fit your search.

Fli is also available as a 1:1 TypeScript port, published to npm as fli-js. Same models, same filter encoding, same direct-API approach.

bun add fli-js   # or: npm install fli-js

import { Airport, FlightSearchFilters, FlightSegment, SearchFlights, SeatType } from "fli-js";

const inDays = (n: number) => new Date(Date.now() + n * 86_400_000).toISOString().slice(0, 10);

const filters = new FlightSearchFilters({
  passenger_info: { adults: 1, children: 0, infants_in_seat: 0, infants_on_lap: 0 },
  flight_segments: [
    new FlightSegment({
      departure_airport: [[[Airport.JFK, 0]]],
      arrival_airport: [[[Airport.LAX, 0]]],
      travel_date: inDays(30),
    }),
  ],
  seat_type: SeatType.ECONOMY,
});

const results = await new SearchFlights().search(filters, { currency: "USD" });

fli-js uses the same search-page transport as the Python package (see Search transport above), with the same consequences: emissions / bags / exclude_basic_economy are dropped with a warning, multi-city throws SearchUnsupportedError, getBookingOptions throws SearchRejectedError, date searches cost one page fetch per date and are capped at 93, and FLI_SOCS_COOKIE controls the consent cookie.

The TypeScript source lives in fli-js/; see the TypeScript Quick Start for the full guide.

# Clone the repository
git clone https://github.com/punitarani/fli.git
cd fli

# Install dependencies with uv
uv sync --all-extras

# Run tests (offline only — no network access required)
uv run pytest

# Run linting
uv run ruff check .
uv run ruff format .

# Build documentation
uv run mkdocs serve

# Or use the Makefile for common tasks
make install-all  # Install all dependencies
make test         # Run tests (offline only)
make test-live    # Run the small, stable live tests (real network)
make lint         # Check code style
make format       # Format code

Tests that call the real Google Flights API are marked live and skipped by default (--fuzz, --live and --all are independent, skip-only gates — --all does not imply --live, and --live alone does not imply --fuzz). Run the small, stable live set with make test-live (or pytest -m live --live); the noisier 100-case fuzz-gated live test is a separate opt-in, make test-live-fuzz (pytest --all -m live --live tests/search/test_search_flights_fuzz.py — it needs --all/--fuzz together with --live, since it carries both markers).

# Build the devcontainer
docker build -t fli-dev -f .devcontainer/Dockerfile .

# Run CI inside the container
docker run --rm fli-dev make lint test-all

# Or run lint and tests separately
docker run --rm fli-dev make lint
docker run --rm fli-dev make test-all

To run GitHub Actions locally, install act:

brew install act

# Run CI locally (lint + tests on Python 3.10-3.13)
make ci

# Or run CI inside Docker (no local act installation needed)
make ci-docker

Contributions are welcome! Please feel free to submit a Pull Request.

This project is licensed under the MIT License — see the LICENSE file for details.