That sounds small until a model summarises three listings and you have no way to find them again. Responses are deliberately flattened and pruned to keep the context cost down, which matters when a search returns a page of properties. Robots rules are followed by default and the override is explicit at both the call and the launch level — which is the right way round.
A search and detail server for Airbnb listings. It parses the site's HTML with cheerio, returns structured JSON, and includes a direct URL on every listing so a result can be opened rather than searched for again.
- Listings searched by `location`, with optional `placeId`, `checkin` and `checkout` dates, guest counts split across `adults`, `children`, `infants` and `pets`, a `minPrice` and `maxPrice` band, and a `cursor` for paging — each result carrying name, price, location and a direct `url` — `airbnb_search`
- One listing read in full by `id` — description, host details, amenities and pricing, with the same date and guest parameters applied and a direct `url` returned — `airbnb_listing_details`
- Responses flattened and trimmed to keep the context cost down rather than returning raw page data
- Airbnb's robots.txt rules respected by default, with `ignoreRobotsText` available per call and `--ignore-robots-txt` as a launch flag
No API key, and Node.js for the build. The README's client entry names `@Domoteek/mcp-server-airbnb` at 0.1.0, which npm does not have, though the project is also listed on Smithery. Building from source is `npm install` and `npm run build`. Domoteek is not affiliated with Airbnb, Inc. or its subsidiaries.
One command — npx -y @smithery/cli install @Domoteek/mcp-server-airbnb --client claude