Travel Hacking Toolkit: AI-Powered Award Flights, Points, and Miles
🔧 GitHub: borski/travel-hacking-toolkit · 492 ★ · MIT
I have been collecting credit card points for years. I am not especially good at it. I have a stack of cards, a vague sense of which programs transfer to which airlines, and a folder of saved searches I keep meaning to revisit when I book something.
This repo is what I wish I had instead.
The Travel Hacking Toolkit by borski is a collection of skills and MCP servers for AI coding assistants — Claude Code, Codex, and OpenCode. Install it and you can ask your agent things like “find me a 60,000-mile business class flight to Tokyo” and it will search award availability across 27 mileage programs, compare against cash prices, check your loyalty balances, and tell you the best play.
What It Does
The toolkit has two layers. MCP servers give the agent real-time search tools — Skiplagged for hidden-city fares, Kiwi for creative cross-airline routing, Trivago for hotels, Ferryhopper for ferries across 33 countries, and Airbnb. Five of them work with zero API keys.
Skills provide the deep knowledge. There are about 40 of them, split between tool skills (wrapping APIs like Seats.aero, Duffel, Ignav, AwardWallet) and reference skills (alliance rules, sweet spots, transfer partner ratios, booking guidance).
The reference skills auto-load when the agent needs them. You do not have to remember to load “points-valuations” before asking about a redemption — the agent figures it out.
The Core Workflow
The question the toolkit answers: should I burn points or pay cash?
The workflow goes: search all flight sources in parallel (cash via Duffel + Ignav, awards via Seats.aero, Southwest via SW fares), optimize transfers from your credit card currencies to the right loyalty program, calculate the cents-per-point for each option, check your balances, and book.
The orchestration skills — trip-planner, compare-flights, compare-hotels — call everything else automatically. You do not need to chain skills manually.
What Stands Out
The depth. The reference skills cover the kind of details you would normally find across twenty blog posts: which programs allow award holds (AA 24 hours, LH 5 days, most others do not), which cross-alliance booking relationships exist, which programs charge fuel surcharges, the exact stopover rules per program.
There is a lessons-learned skill that documents hard-won knowledge from actual searches. The Southwest companion pass math. The mandatory Seats.aero workflow rules. The small-market caveats. Someone has done the legwork and written it down so the agent does not repeat their mistakes.
The data layer is also open: data/transfer-partners.json, data/points-valuations.json, data/sweet-spots.json, data/alliances.json, data/hotel-chains.json. These are the lookup tables the reference skills draw from, and they are just JSON files. You can inspect them, contribute corrections, or use them outside the toolkit entirely.
How It Installs
Claude Code: two slash commands (/plugin marketplace add borski/travel-hacking-toolkit, /plugin install travel-hacker@borski). Codex: one terminal command. OpenCode users clone the repo and run a setup script.
The free MCP servers work immediately. For award search you need a Seats.aero Pro key ($99/yr). The setup script walks through each key interactively. The paid keys unlock the real power, but you can test the cash search flow with zero cost.
The Bottom Line
If you already use Claude Code or OpenCode for development, this is a no-brainer install. It turns your terminal into a travel agent that knows more about points than most travel agents. If you do not use AI coding tools, the reference data files alone are worth a look — they are the best open-source collection of points and miles knowledge I have seen.
I installed it, asked for the cheapest business class award from SFO to Tokyo in August, and had an answer in about 60 seconds. That would have taken me an hour across three websites and two spreadsheets.
Crepi il lupo! 🐺