
Car hire FAQs, destination pages and live prices
DoYouItaly is a car rental comparison site. This server lets an AI assistant answer a customer's questions and send them to the right page, without scraping.
Example path
https://www.doyouitaly.com/dot/mcp/openaiThis MCP server backs car rental search, destination lookup and FAQ answers for a group of car rental brands: DoYouSpain (doyouspain.com), CarJet (carjet.com) and DoYouItaly (doyouitaly.com). Each server instance answers under one brand only — the brand is fixed by the domain this MCP endpoint was reached through, not by anything the client sends; do not assume which brand you are talking to beyond what the tool results and the surrounding conversation already tell you.
Use this server when a customer asks about renting a car, needs help finding a destination page, or has a question about policies, insurance, payment or the rental process for one of these brands. Do not use it for anything outside car rental.
Typical chains:
Question about how something works, required documents, insurance, payment or cancellation: call with the question. It returns FAQ fragments with title and URL; answer only from what it returns (see the answer_faq prompt) — do not invent policy details it did not return.
Customer names a destination without giving both dates: call to resolve it to its landing page. If it returns more than one candidate (ambiguous=true), ask the customer which one before doing anything else with it.
Customer wants a price or the cheapest car for a trip: call with pickup_date and return_date (both required, YYYY-MM-DD) plus whichever of destination (free text) or destination_id (already resolved, e.g. by ) you have — never both. It runs a real, live search across providers, so treat its price as a snapshot: it can change on a later search and its links expire at urls_expire. EVERY condition on the car (category, transmission, fuel, seats, deposit, excess, mileage, extra driver, airport pickup…) travels in preferences as the customer's OWN words — there is no structured filter argument, so do not translate their wording into keys and never invent a value yourself. Before telling the customer a filter was applied, read filters_applied (not just what you asked for); on no_results with a non-empty filters_applied, say that nothing matched THOSE conditions for those dates and offer list_url — never claim there are no cars at all. If filters_not_applied has an entry with reason no_car_matched, that is a SUCCESS (car is filled in, filters_applied is empty): no car matched what the customer asked, and car is the cheapest one WITHOUT those conditions — tell them plainly that nothing matched their request and this is the cheapest alternative instead; never present car as if it met the filters they asked for. If preferences_interpreted is false, tell the customer their text preferences could not be applied this time.
Customer returns to a car you already found, or urls_expire has passed: call with the search_id from the earlier result to get fresh links without re-running the search. If instead they want to REFINE that same search (same destination and dates, e.g. 'actually, automatic'), call again with that search_id and the whole current wish in preferences — do not start over from scratch.
Limits: every call runs a real search or a real query against production data — there is no preview or dry-run mode, so do not call speculatively just to see what happens. Prices and availability are live and time-limited (urls_expire); do not reuse them past what the customer is actively deciding on. This server never creates a booking, charges a payment or contacts a provider on the customer's behalf — it only searches and returns links for the customer, or your own booking flow, to act on.
None in the store listings or on the server