Labsco
MCP SERVER

AirTreks MCP

by SEKeener

Plan a multi-stop or round-the-world trip and find out which routings will actually ticket — before anyone tries to buy one.

Travel Booking, Transit & Navigation
Summary
The valuable answer is which routing fails, and it arrives before anyone spends money.

Alliance round-the-world fares can be assembled on a booking site and then collapse at ticketing, and this server puts a measured number on that risk: 91% bookability at three or four legs, 61% at five or six, 6% at seven or more, from real fare-construction attempts. What comes back alongside the verdict is the alternative — a carrier for each leg, with the reason, and the sectors better travelled overland. Use route_estimate for the early question, since it answers from fare history in about a quarter of a second with no dates at all, and save itinerary_quote for a trip that is real: it takes about a minute and returns a reference to poll. trip_idea_create submits a customer's actual contact details, so only call it with their knowledge and consent.

What it is

A hosted MCP server for AirTreks at https://mcp.airtreks.com/mcp, aimed at the itineraries ordinary flight search gives up on: multi-stop, round-the-world, open-jaw and surface segments. Give it a list of cities and it weighs Star Alliance and oneworld round-the-world fares against a custom mixed-carrier build, then says which one holds together at ticketing. Free, no key, 100 requests a day per IP.

What you get
  • plan_route is the entry point: cities in, and back comes a per-leg carrier plan with alternatives, the sectors worth covering overland instead of flying, and a recommended approach with the reasoning attached
  • route_validate checks a routing against alliance rules, known dead legs, poison carriers and bookability; hub_check finds the working connection between two airports
  • route_suggest offers three routings by region, direction and alliance, and custom_route_build breaks a complex itinerary into individually ticketable segments
  • fare_product_match names the fare product that fits — RTW, Circle Pacific, Circle Atlantic, open jaw or custom — with its typical price range
  • route_estimate returns a price range from AirTreks fare history in about a quarter of a second and needs no dates, which suits the question asked before anything is decided
  • fare_quote prices one specific itinerary on specific dates as a single ticket; itinerary_quote prices a whole multi-stop trip several ways — cheapest, fastest, fewest stops — and shows which tickets each version is built from
  • itinerary_quote takes about a minute, so it returns a reference immediately and itinerary_quote_status fetches the result, free to poll
  • trip_idea_create hands the full routing analysis to an AirTreks consultant who prices and books the real ticket
  • The bookability rates behind those recommendations are measured rather than estimated: 91% at 3-4 legs, 61% at 5-6, and 6% at 7 or more
Requirements

Point a client at https://mcp.airtreks.com/mcp over streamable HTTP. No API key for any tool, the consultant handoff included, at 100 requests a day per IP; a key raises that limit and travels in an X-API-Key header, or as ?key= for clients that cannot set headers, as claude.ai connectors cannot. npx airtreks-mcp runs the same server locally over stdio. Every tool also exists as a REST endpoint at /api/{tool}, with an OpenAPI 3.1 spec at /openapi.json. AGPL-3.0-only.

Setup effort

One command — npx airtreks-mcp