Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
jackculpan avatar

Flightclaw

  • 1k installs
  • 65 repo stars
  • Updated June 6, 2026
  • jackculpan/flightclaw

flightclaw is a Python agent skill and local MCP server that searches Google Flights, finds cheapest travel dates, and tracks route prices for developers who need programmatic fare research and drop alerts.

About

flightclaw is a Python agent skill and local MCP server that queries Google Flights through the flights pip library to search routes, surface cheapest dates across ranges, and monitor fares over time. The project exposes 6 MCP tools—search_flights, search_dates, track_flight, check_prices, list_tracked, and remove_tracked—and ships 4 CLI scripts under scripts/ for terminal workflows. Search filters cover passengers, airlines, USD price caps, duration, departure and arrival times, layover limits, multi-airport codes, date ranges, and 6 sort modes from BEST to DURATION. Price history persists in data/tracked.json with optional target-price alerts, and returned fares use the user's auto-detected local currency. Install with npx skills add jackculpan/flightclaw or pip install flights mcp[cli] on Python 3.10+, then wire server.py into Claude Code or any MCP client. Reach for flightclaw when planning conference travel, comparing relocation routes, or estimating location-based product costs inside an agent session.

  • Search Google Flights data for routes, dates, airlines, cabin classes, stops and price filters
  • Track price changes for specific routes over time with historical monitoring
  • Set up price-drop alerts for chosen routes and travel parameters
  • Runs as both a CLI tool and an MCP server
  • Supports multi-airport searches and flexible date ranges in one command

Flightclaw by the numbers

  • 1,030 all-time installs (skills.sh)
  • +6 installs in the week ending Aug 5, 2026 (Skillselion tracking)
  • Ranked #466 of 3,282 Productivity & Planning skills by installs in the Skillselion catalog
  • Security screen: MEDIUM risk (skills.sh audit)
  • Data as of Aug 5, 2026 (Skillselion catalog sync)
npx skills add https://github.com/jackculpan/flightclaw --skill flightclaw

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Installs1k
repo stars65
Security audit1 / 3 scanners passed
Last updatedJune 6, 2026
Repositoryjackculpan/flightclaw

How do you track Google Flights prices with MCP?

Quickly research real flight prices, cheapest dates, and route trends when planning travel, events, or location-based product ideas.

Who is it for?

Developers wiring Claude Code or MCP clients who need programmatic Google Flights search, cheapest-date discovery, and ongoing route price monitoring.

Skip if: Teams needing airline booking APIs, corporate travel management, or production fare-commerce integrations instead of personal research tooling.

When should I use this skill?

A developer asks to search flights, find cheapest travel dates, add a route to tracking, or check tracked fares for price drops.

What you get

Route search results, cheapest-date calendars, data/tracked.json price history, and target-price drop alerts

  • data/tracked.json price history
  • flight search results
  • price-drop alerts

By the numbers

  • Exposes 6 MCP tools for search, date lookup, tracking, and price alerts
  • Ships 4 CLI scripts in scripts/ for search-flights, track-flight, check-prices, and list-tracked
  • Requires Python 3.10+ and 2 pip packages: flights and mcp[cli]

Files

SKILL.mdMarkdownGitHub ↗

flightclaw

FlightClaw is a personal travel-booking agent. It remembers who you are, who you travel with, the loyalty programs and cards you hold, and how you like to fly — then recommends, books, pays for, and learns from each trip.

Personalization data (travelers, preferences, cards/points, companion groups, trip history) is stored server-side in the private flightclaw-api Worker (D1), reached via FLIGHTCLAW_API_URL + FLIGHTCLAW_API_KEY. Payment for bookings routes through Link virtual cards (duffel_book_with_link).

The flow

1. Onboarding (one time)

Set this up once, then reuse forever.

1. Who you aresave_traveler for yourself (full passport-accurate name, DOB, contact, loyalty programmes), then set_me to mark that profile as you and record home airports. 2. Companionssave_traveler for each person you travel with (relationship: spouse/partner/child/parent/friend/colleague). Group them with save_group (e.g. family = jack,jane). 3. Preferencesset_preferences: cabin by haul (e.g. short-haul ECONOMY, long-haul BUSINESS), preferred/avoided airlines, alliance, seat, departure window, max stops, red-eye tolerance, baggage, meal, and budget sensitivity (cheapest / balanced / comfort). 4. Cards & pointssave_card for each card; set_points_balance for each loyalty/transfer program. Then enrich with the Card Links MCP (find_transfer_programs_for_airline, list_transfer_partners) so you know which airlines each card's points can reach.

2. Planning a trip

1. Ask where they want to go and who's coming — reuse a group with get_group (it returns the exact passengers string for booking) or make a new one with save_group. 2. Recommendrecommend_flights(origin, destination, date, ...). It loads the saved preferences, picks the cabin by haul, drops avoided airlines, and ranks options on price/duration/stops/preferred-airline/departure-window/ red-eye, returning the top 3 with a "why this fits you" for each. 3. Awards / points option — if they want to spend points, call the Award Travel Finder MCP (search_availability, search_all_airlines, get_pricing) using their stored loyalty programs and points balances, and present award options alongside the cash fares ("best overall / cheapest / best points value").

3. Booking & paying

1. Get a bookable, payable offer with duffel_search_flights (real fares/ conditions). duffel_get_offer / duffel_get_seat_map for extras. 2. Confirm the choice with the user, then `duffel_book_with_link` with the group's passengers string. This creates a Link spend request (the user approves the charge, ≤ $500), returns a virtual card + Duffel checkout URL, and you complete payment via Chrome automation. For higher amounts use duffel_book_flight (Duffel balance) or duffel_create_checkout. 3. `log_trip` right after booking (route, dates, travelers, cabin, price, order_id) so it enters history and the follow-up queue.

4. Post-trip follow-up & learning (the real magic)

1. trips_pending_followup surfaces trips that have completed/returned. 2. Ask how each went, then record_trip_feedback(id, feedback, learnings=...). Durable lessons (e.g. "prefers window on long-haul", "dislikes early departures") are appended to the user's preferences, so the next recommend_flights is sharper. Over time FlightClaw learns the traveler.

Tools

Personalization (backend-backed)

  • Travelers: save_traveler, list_travelers, get_traveler, delete_traveler,

set_me, get_me, import_local_passengers (one-time migration of any old local data/passengers.json).

  • Preferences: set_preferences, get_preferences, update_preferences.
  • Cards/points: save_card, list_cards, delete_card, set_points_balance,

list_points.

  • Groups: save_group, list_groups, get_group, delete_group.
  • Trips: log_trip, list_trips, get_trip, trips_pending_followup,

record_trip_feedback.

  • Recommendation: recommend_flights.

Search & trackingsearch_flights, search_dates, track_flight, check_prices, list_tracked, remove_tracked.

Booking (Duffel)duffel_search_flights, duffel_search_multi_city, duffel_get_offer, duffel_get_seat_map, duffel_book_flight, duffel_book_with_link, duffel_create_checkout, duffel_list_orders, duffel_get_order, duffel_request_change, duffel_confirm_change, duffel_cancel_order, duffel_confirm_cancel, duffel_check_alerts, link_list_payment_methods.

External MCP integration

FlightClaw stores the user's cards/points; the agent enriches and acts on them using two other MCP servers when present:

  • Card Links — transfer partners and card comparisons for the user's stored

cards.

  • Award Travel Finder — award availability and points pricing across

airlines/programs.

When surfacing card recommendations from Card Links, always include its disclaimers: not financial advice; affiliate links may earn commission; card terms change — verify current offers with the issuer.

Setup

pip install flights "mcp[cli]"
export FLIGHTCLAW_API_URL="https://flightclaw-api.<your>.workers.dev"
export FLIGHTCLAW_API_KEY="<your API key>"
claude mcp add flightclaw -- python3 /path/to/flightclaw/server.py

The Worker (flightclaw-api) holds the Duffel token and D1 profile store; apply schema.sql once with wrangler d1 execute flightclaw-db --remote --file schema.sql.

Data

Personalization data is server-side (D1). Price-tracking history (data/tracked.json) and a local Duffel order cache (data/duffel_orders.json) remain local and are gitignored.

Related skills

How it compares

Pick flightclaw for agent-driven Google Flights research and ongoing route tracking; use dedicated travel booking apps when you only need a one-off purchase UI.

FAQ

How does flightclaw get flight prices?

flightclaw queries Google Flights through the Python flights pip library. Search and tracking tools return fares in the user's local currency auto-detected from IP, with ongoing history stored in data/tracked.json.

What MCP tools does flightclaw expose?

flightclaw registers 6 MCP tools: search_flights, search_dates, track_flight, check_prices, list_tracked, and remove_tracked. Claude Code or any MCP client can call them after pip install flights mcp[cli].

Can flightclaw search multiple airports and date ranges?

flightclaw accepts comma-separated airport codes like LHR,MAN and date_to ranges that expand into per-day searches. Filters include airline, max USD price, duration, layovers, and passenger counts.

Is Flightclaw safe to install?

skills.sh reports 1 of 3 security scanners passed. Review the Security Audits panel on this page before installing in production.

Productivity & Planningautomationresearch

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.