Someone built a travel hacking toolkit that actually works with AI agents. The open-source project plugs into Claude Code and OpenCode via MCP servers, letting you ask your AI assistant to find award flights across 25+ mileage programs, compare points versus cash prices, and check your loyalty balances. Five MCP servers (Skiplagged, Kiwi, Trivago, Ferryhopper, Airbnb) work immediately with zero API keys. The premium features need Seats.aero for award searches (~$8/month) and SerpAPI for cash price comparisons.
The core workflow answers the one question every travel hacker asks: "Should I burn points or pay cash?" Your AI agent searches award availability, pulls cash rates from Google Flights via SerpAPI, calculates the cents-per-point value, checks your balances through AwardWallet, and tells you the best move. The toolkit includes data files with alliance memberships, transfer partners, and sweet-spot award routes so the agent actually knows what it's doing.
Most agent demos feel like parlor tricks. This one solves a genuinely annoying problem. Award search is fragmented across dozens of airline sites. Cash prices bounce constantly. The math takes time even for experienced travel hackers. MCP architecture means the AI can call travel APIs directly without you babysitting each step. Hacker News commenters pointed out real gaps, like finding 3+ award seats for families or integrating with credit card hotel portals. For solo travelers with points scattered across programs, this beats checking United or American manually.