EXTENSIONS
Ignav
Flight fares from Ignav, for answering "which flight gets me there in time": search one-way legs by route and date, filtered by airline, nonstop and departure/arrival hour windows; fan out over every outbound and return candidate of a trip in one execution; and get the airline's booking link for the flight you pick. Times are stored local to each airport with timezone and UTC.
Trip Planner
Trip-planner services with canned data, for workshops and demos: extract (structured trip requirements from a plain-English request, standing in for a Claude Haiku call), search_flights (priced flight options, standing in for Google Flights) and search_hotels (AI-rated hotels under a nightly cap, standing in for hotelist.com). Same arguments and resource shapes as @adam/trip-planner and @keeb/hotelist, so an itinerary behavior and a plan-trip workflow built on these run with no API keys and no network. The data is fixed: a comfortable 10-day April trip from SFO to Tokyo on a $4,000 budget, three flights and four hotels. Built for the Behavior-Centric AI Design workshop.
Ns International
Plan train trips on NS and NS International via the official NS
Trip Planner
Plan a trip from one plain-English sentence. Claude Haiku extracts the
Hotelist
AI-rated hotel search via the hotelist.com MCP server. Wraps all three of the server's tools as model methods: search_hotels (city shortlist, max 10, with 0-10 AI rating computed from reviews and photos, nightly USD price, and AI-summarized pros/cons), get_hotel (full detail — description, per-dimension rating breakdown, amenities, wifi speed, photo), and list_cities (coverage with hotel counts and average rating). The endpoint is a stateless streamable-HTTP MCP server needing no initialize handshake, session header, or authentication. Ratings are not pay-to-play. Includes a bundled hotel-recommendation workflow that shortlists a city, asserts at least one result is actually in that city (the API also returns nearby ones), and pulls full detail on the top in-city pick.