Key Takeaways
- The Travel Hacking Toolkit is an open-source project that connects AI coding assistants to flight, hotel, and ferry search APIs through MCP servers and drop-in skills.
- Five MCP servers (Skiplagged, Kiwi.com, Trivago, Ferryhopper, Airbnb) work immediately with zero API keys.
- Award flight search across 25+ mileage programs requires a Seats.aero Pro subscription at approximately $8/month.
- The toolkit supports both Claude Code and OpenCode, with a guided setup script that configures API keys and installs dependencies automatically.
What Happened
A developer named Borski released the Travel Hacking Toolkit on GitHub on April 4, 2026, providing a set of MCP servers and skills that let AI coding assistants search for flights, hotels, and ferry routes. The project appeared as a Show HN post and reached 65 points on Hacker News.
The toolkit turns Claude Code or OpenCode into a travel planning agent. Users can ask their AI assistant to “find a 60,000-mile business class flight to Tokyo,” and the system searches award availability across 25+ loyalty programs, compares against cash prices, checks loyalty balances, and recommends the best option.
Why It Matters
MCP (Model Context Protocol) servers are becoming the standard way to extend AI assistants with real-world data access. The Travel Hacking Toolkit demonstrates a practical, consumer-facing use case: turning the complexity of award travel optimization into a natural language conversation. Award flight search typically requires checking multiple programs individually, comparing redemption rates, and factoring in transfer partners. Automating this through an AI agent eliminates hours of manual searching.
The toolkit also showcases the growing ecosystem around Claude Code’s MCP integration, with the project using the --strict-mcp-config flag for reliable server loading from configuration files. This pattern — wrapping existing travel APIs into MCP-compatible servers — is replicable for virtually any API-driven domain or service category.
Technical Details
Installation runs through a single setup script: git clone the repository and run ./scripts/setup.sh. The script detects which AI tool the user prefers, creates API key configuration files, installs dependencies, and optionally installs skills system-wide.
The five free MCP servers provide immediate functionality without any API keys:
- Skiplagged – Flight search with hidden city fares, a technique that exploits cheaper routes through connection cities
- Kiwi.com – Flights with virtual interlining for creative cross-airline routing that traditional booking engines cannot construct
- Trivago – Hotel metasearch across major booking sites for price comparison
- Ferryhopper – Ferry routes across 33 countries and 190+ operators, covering Europe, Asia, and beyond
- Airbnb – Listing search with a patched geocoding fix and property type filter added by the toolkit developer
For full award travel functionality, two paid API keys are recommended: Seats.aero (Pro tier at ~$8/month) for award flight availability across 25+ mileage programs, and SerpAPI (free tier with 100 searches/month) for Google Flights cash price comparisons that help determine whether redeeming points or paying cash yields better value.
Additional skills include Duffel for real-time airline search across carriers, AwardWallet for tracking loyalty balances and elite status across programs, and LiteAPI for hotel booking with live rates. For Claude Code, the recommended launch command is claude --strict-mcp-config --mcp-config .mcp.json, which the developer notes is more reliable than auto-discovery for MCP server loading.
Who’s Affected
Frequent travelers who actively manage points and miles across multiple loyalty programs stand to gain the most. The toolkit reduces the manual work of checking award availability across airlines and transfer partners — a process that can take hours per trip when done manually across individual airline websites.
Developers interested in building MCP servers will find the project useful as a reference implementation showing how to wrap third-party APIs into the MCP format. Claude Code and OpenCode users gain a new category of practical tooling beyond software development, expanding what AI coding assistants can do in daily workflows.
What’s Next
The project is open-source under active development on GitHub. As MCP adoption grows across AI assistants, similar toolkits for other domains — personal finance, healthcare appointment booking, real estate search — are likely to follow the same pattern of wrapping existing APIs into AI-accessible MCP servers. The travel hacking use case is particularly compelling because it combines multiple data sources that no single existing tool aggregates well.
