Connect an AI assistant (MCP)
Add https://worththestay.com/mcp as a custom connector in Claude, ChatGPT or any MCP client. It speaks streamable HTTP (POST JSON-RPC 2.0, stateless).
| Tool | What it returns |
|---|---|
search_hotels | best hotels in a city on the corrected score, with margin, evidence, pros/cons, page URL, booking_links |
get_hotel | one hotel with the working behind its score, description, amenities, better alternatives nearby, booking_links |
list_cities | covered cities with hotel counts, average score and page URL |
Or call it directly:
curl -s -X POST https://worththestay.com/mcp -H 'Content-Type: application/json' \
-d '{"jsonrpc":"2.0","id":1,"method":"tools/call","params":{"name":"search_hotels","arguments":{"city":"Paris"}}}'
Limits per IP: 30 requests/minute, 400/day, 50 full hotel details/day.
JSON endpoints
| Request | What it returns |
|---|---|
GET /api/docs | this document |
GET /api/ts/search?city=<city>&limit=24 | hotels in a city, best corrected score first; <city> is the slug from its page URL (cape-town-za) or its name (Cape Town). Also q=<text>, offset=<n> |
GET /api/ts/hotel/<slug> | one hotel, as the site shows it |
GET /sitemap.xml | sitemap index: screens, countries, cities, hotels, guides |
Try them: hotels in Paris · Maison Villeroy · this page as JSON
What the score means
Raw review scores bunch near the top, so each platform is rescaled against its own 5th-percentile floor in our corpus. That platform part (60%) is weighed with an AI score of the photo set (25%) and an AI reading of review text (15%), and pulled toward the corpus average when few platforms rate the hotel or they disagree. Scale: 0-5, one decimal, shown with a +/- margin.
| Field | Meaning |
|---|---|
score | the corrected score |
score_margin | the +/- |
evidence | solid, fair or thin |
platforms_used | of Booking.com, Tripadvisor, Google, Agoda |
platform_average | the plain mean before correction |
The full working is on the method page.
Good to know
- No hotel can pay for placement or a better score.
- Prices are a cached snapshot, not a live quote. Use booking_links for the real rate.
- booking_links are ready-made per-site booking URLs. Use them verbatim; some may fund this site at no cost to the user.
- Hotel data from OpenStreetMap contributors (ODbL), merged with a commercial hotel catalogue for prices, photos and platform ratings. Link back when you use it.
For language models there is also llms.txt.