Skip to main content
Glama
deenesjakoruzh
mtnrabi

google-flights-mcp

by mtnrabi
OverviewSchemaSourceRelated ServersDiscussionsAdmin
Python
Hybrid

Google Flights MCP — тарифы в реальном времени, которые ваш агент может искать по всему диапазону дат, без рекламы

claude mcp add --transport http google-flights https://google-flights-mcp.flightpowers.com/mcp --header "x-rapidapi-key: YOUR_RAPIDAPI_KEY"

Хостится. Ничего клонировать не нужно, ничего собирать не нужно. Включён в официальный реестр MCP Registry как com.flightpowers/google-flights-mcp. Проверка состояния: /health.

Нужен ключ? Подпишитесь на Google Flights Live API на RapidAPI — доступен бесплатный тариф — и скопируйте свой x-rapidapi-key: https://rapidapi.com/mtnrabi/api/google-flights-live-api

Пока нет ключа? Начните с бесплатного сервера — тот же поиск, без регистрации: claude mcp add --transport http google-flights-free https://google-flights-lulu.flightpowers.com/mcp (на рекламе: одна раскрытая спонсорская карточка на результат, разветвление ограничено 15, а клиенты, которые не могут отображать спонсорскую карточку, могут быть ограничены дополнительно.) Возвращайтесь сюда, когда реклама, лимит в 15 поисков или ограничения клиента начнут вам мешать.

Что получает ваш агент

Два инструмента отвечают на вопрос о цене, а не на поиск по дате.

  • Задавайте открытые вопросы. «Самый дешёвый билет в одну сторону до Шри-Ланки в любой день октября», «от 5 до 7 ночей в Риме когда-нибудь в мае, из Тель-Авива или Ларнаки» — каждый такой запрос — это один вызов инструмента. Оба инструмента принимают диапазон дат вылета, список аэропортов назначения и (для перелёта туда и обратно) значение nights вместо фиксированной даты возвращения и разворачивают их внутренне.

  • Сообщайте, действительно ли цена хорошая. Каждый результат содержит собственный исторический диапазон цен Google для этого маршрута и периода — price_insights_low, price_insights_high и вердикт price_range_in_relation_to_other_periods со значением low / typical / high. Это позволяет агенту ответить «$209 — это обычная цена, не торопитесь», а не просто назвать число.

  • Бронируйте, а не только смотрите. Каждый результат содержит buy_link на Google Flights.

  • Отслеживайте расходы. Каждый ответ содержит api_usage — запросы, использованные этим вызовом, и что осталось на плане вызывающего. См. Отчёт о расходах.

  • Отслеживайте результаты поиска. Каждый ответ содержит search_coverage, чтобы модель могла честно сказать, на каких датах и направлениях основан ответ.

Результаты — это тарифы в реальном времени. Они устаревают в течение минут — никогда не кэшируйте тариф и не переиспользуйте более ранний результат; ищите заново и указывайте, когда были получены данные.

Related MCP server: Ignav Flights MCP Server

Получение ключа (доступен бесплатный тариф)

Сервер не хранит собственных внешних учётных данных. Каждый поиск оплачивается из вашей подписки RapidAPI — поэтому ключ передаётся вместе с запросом.

  1. Подпишитесь на Google Flights Live API: https://rapidapi.com/mtnrabi/api/google-flights-live-api

  2. Скопируйте свой x-rapidapi-key.

  3. Передайте его серверу любым из трёх способов ниже.

Если ключ отсутствует, инструменты не завершаются с ошибкой молча и ничего не тратят — они возвращают needs_api_key: true вместе с URL для регистрации и этими инструкциями, сформулированными так, чтобы модель могла зачитать их вам.

Три способа передать ключ

Способ

Как

Когда использовать

Заголовок (предпочтительно)

--header "x-rapidapi-key: YOUR_RAPIDAPI_KEY"

Всё, что позволяет задавать заголовки. Ключи не попадают в URL, а значит, не попадают в журналы прокси и журналы доступа.

Параметр запроса

https://google-flights-mcp.flightpowers.com/mcp?rapidapi_key=YOUR_RAPIDAPI_KEY

Хосты, которым можно вставить только URL, — в первую очередь диалог настраиваемого коннектора в claude.ai.

Поле API-ключа клиента

Вставьте ключ в собственное поле «API key» клиента

Хосты, которые отправляют authorization: Bearer <key> или x-api-key. Также принимается форма сохранённой конфигурации Smithery (config.rapidApiKey=).

Приоритет у первого непустого источника, в этом порядке. Ключ никогда не логируется, не повторяется в сообщении об ошибке и не возвращается в ответе инструмента.

Инструменты

Инструмент

Что делает

search_oneway_flights

Тарифы в одну сторону в реальном времени. Исходные данные: IATA-код отправления, IATA-код назначения или список, и либо одна дата вылета, либо диапазон дат. Возвращает цену, авиакомпанию, длительность, стопы, buy_link и ствой исторический диапазон цен Google, чтобы можно было оценить тариф. Используйте для любых вопросов про билет в одну сторону, включая открытые: один вызов с диапазоном, а не один вызов на дату.

search_roundtrip_flights

Тарифы туда и обратно в реальном времени, рассчитанные как парные сегменты, а не два билета в одну сторону. Исходные данные: отправление, назначение(я), дата отправления или диапазон, и либо return_date, либо длина поездки в nights (число или список, например [5,6,7]). Возвращает общую цену, по каждому сегменту авиакомпанию/стопы/длительность и один buy_link для поездки.

search_oneway_flights

search_oneway_flights(
    from_airport: str,                     # origin IATA, e.g. "TLV"
    to_airport: str | list[str],           # destination IATA, or a list to compare
    departure_date: str | None = None,     # "YYYY-MM-DD"
    departure_date_from: str | None = None,# first date of a range
    departure_date_to: str | None = None,  # last date of a range
    max_stops: int | None = None,          # 0 = non-stop only
    airline_codes: list[str] | None = None,
    exclude_airline_codes: list[str] | None = None,
    departure_time_min: int | None = None, # hour, 0-23
    departure_time_max: int | None = None,
    arrival_time_min: int | None = None,
    arrival_time_max: int | None = None,
    currency: str = "usd",
    max_price: int | None = None,
    seat_type: int | None = None,          # 1 economy, 2 premium economy, 3 business, 4 first
    passengers: list[int] | None = None,   # [adults, children, infants]
    sort_by: str = "best",                 # "best" | "price" | "duration"
    limit: int = 10,                       # results returned after merge + sort
    max_searches: int | None = None,       # cap the billed requests this call may make
    use_fallback: bool = False,            # slower, fewer empty results on hard routes
)

search_roundtrip_flights

search_roundtrip_flights(
    from_airport: str,
    to_airport: str | list[str],
    departure_date: str | None = None,
    departure_date_from: str | None = None,
    departure_date_to: str | None = None,
    return_date: str | None = None,        # use this OR nights, not both
    nights: int | list[int] | None = None, # e.g. 7, or [5, 6, 7]
    max_departure_stops: int | None = None,
    max_return_stops: int | None = None,
    departure_airline_codes: list[str] | None = None,
    return_airline_codes: list[str] | None = None,
    currency: str = "usd",
    max_price: int | None = None,
    seat_type: int | None = None,
    passengers: list[int] | None = None,
    sort_by: str = "best",
    limit: int = 10,
    max_searches: int | None = None,
    use_fallback: bool = False,
)

sort_by применяется этим сервером к объединённому набору результатов из всех выполненных поисков, поэтому остановка происходит предсказуемо независимо от того, сколько комбинаций было развёрнуто.

Рабочий пример

Пользователь: «Я в Тель-Авиве. Самый дешёвый недельный тур Рим или Афины, вылет в любой день первой половину мая».

Один вызов:

{
  "name": "search_roundtrip_flights",
  "arguments": {
    "from_airport": "TLV",
    "to_airport": ["FCO", "ATH"],
    "departure_date_from": "2026-05-01",
    "departure_date_to": "2026-05-15",
    "nights": 7,
    "sort_by": "price",
    "limit": 5
  }
}

Это разворачивается в 15 дат × 2 направления = 30 комбинаций, что точно равно лимиту на один вызов. Формат ответа (имена полей настоящие; значения ниже приведены для иллюстрации, это не реальный тариф — выполните вызов, чтобы получить живые тарифы):

{
  "results": [
    {
      "from_airport": "Tel Aviv (TLV)",
      "to_airport": "Rome (FCO)",
      "departure_date": "2026-05-05",
      "return_date": "2026-05-12",
      "total_price": "$XXX",
      "total_price_as_number": 0,
      "total_duration_seconds": 0,
      "total_stops": 0,
      "price_range_in_relation_to_other_periods": "low",
      "price_insights_low": 0,
      "price_insights_high": 0,
      "departure_flight_airline": "...",
      "departure_flight_departure_description": "...",
      "departure_flight_arrival_description": "...",
      "departure_flight_duration": "...",
      "departure_flight_stops": 0,
      "departure_stops_info": [],
      "return_flight_airline": "...",
      "return_flight_departure_description": "...",
      "return_flight_arrival_description": "...",
      "return_flight_duration": "...",
      "return_flight_stops": 0,
      "return_stops_info": [],
      "buy_link": "https://www.google.com/travel/flights?tfs=..."
    }
  ],
  "result_count": 5,
  "search_coverage": {
    "requested_combinations": 30,
    "searched_combinations": 30,
    "truncated": false,
    "max_searches_per_request": 30,
    "departure_dates_searched": ["2026-05-01", "..."],
    "destinations_searched": ["ATH", "FCO"]
  },
  "api_usage": {
    "requests_used_by_this_call": 30,
    "plan_requests_remaining": 0,
    "plan_requests_limit": 0,
    "note": "This search used 30 of your RapidAPI plan's requests; ... remain in the current period. Each date and destination combination is one billed request."
  }
}

Другие варианты формы ответа, которые могут встретят за все нормы:

  • На эти даты рейсов нет. results: [] с message — Google Flights действительно не возвращает ничего для некоторых комбинаций маршрута и даты. Это не ошибка. Попробуйте соседние даты, близлежащий аэропорт или use_fallback: true.

  • Часть поисков завершилась ошибкой. Поле partial указывает, сколько из выполненных поисков не прошло, а результаты покрывают остальные.

  • Диапазон слишком широкий. search_coverage.truncated: true и note plus. Диапазон выбирается равномерно по всему окну (первая и последняя дата сохраняются), а не обрезается — поэтому выборка репрезентативна, а не просто первые N дней. Увеличьте max_searches или сузьте диапазон для более полного покрытия.

  • Нет ключа / ключ отклонён. needs_api_key: true, нулевой расход и инструкция по исправлению. Чаще всего влияет ключ RapidAPI, который не подписан именно на этот API.

  • Лимит плана исчерпан. quota_exhausted: true с api_usage, плюс напоминание, что сужение диапазона растягивает оставшуюся квоту.

Отчёт о расходах (api_usage)

Деньги ваши, поэтому счётчик виден. Каждый успешный ответ содержит:

Поле

Значение

requests_used_by_this_call

Сколько платных внешних запросов потребился этот вызов инструмента.

plan_requests_remaining

Сколько осталось на вашем плане RapidAPI в этом периоде.

plan_requests_limit

Лимит вашего тарифа на период.

note

То же самое одним предложением, чтобы модель могла пересказать вам это до того, как вы спрось.

plan_requests_remaining и plan_requests_limit приходят из ответа вышестоящего сервиса и опускаются, если он их не сообщает; note адаптируется соответственно. Правило, которое модель должна проговорить вслух: одна дата × одно направление = один платный запрос.

Рычаги управления расходованием, в порядке грубости: max_searches на вызов (меньше значение — меньше расход на широкий вопрос), более узкий диапазон дат, более короткий список направлений.

Один вызов вместо тридцати

Базовый REST API принимает ровно один кортеж (origin, destination, date) за вызов. При сквозной передаче один вызов на дату «самый дешёвый билет в Шри-Ланку в любой день октября» — это 31 отдельных вызовов инструмента: 31 цикл через модель, 31 шанс потерять ход мысли и счёт, о котором пользователь журнал узнаé после.

Здесь это один вызов инструмента. Развёр поход происходит на стороне сервера, параллельно, с ограничением, равномерной выборкой, дедупликацией на buy_link, объединением и сортировкой по вашему sort_by, и она честно отражается в search_coverage и api_usage.

Этот сервер (платный)

Бесплатный сервер

Разветвление на вызов

30 (жёсткий максимум 60; менять в до callной с помощью max_searches)

15

Реклама

нет

одна раскрытая спонсорская карточка на результат

Ключ

ваш собственный клю RapidAPI

не нежен

Отчёт о расходах

api_usage в каждом ответе

—

Возможность из каталога

да

нет

Это не сервер без рекламы вообще — не из-за вкуса, а из-за ограничений: политика каталога коннек торов Anthropic и правила OpenAI для приложений запрещают рекламу и рекламный контент в результатах работы инструментов, поэтому сервер с рекламой никогда не может быть включён в каталоги, а этот — может.

Локальная разработка

git clone <this repo> && cd mcp_server_paid
python3 -m venv .venv && .venv/bin/pip install -r requirements.txt
cp example.env .env          # fill it in; leave RAPIDAPI_KEY empty
set -a && . .env && set +a
.venv/bin/python -m src      # streamable HTTP on http://localhost:8000/mcp

Подключиёте клиент к локальному процессу так же:

claude mcp add --transport http google-flights-local http://localhost:8000/mcp --header "x-rapidapi-key: YOUR_RAPIDAPI_KEY"

Тесты (124 — проходят, проверено):

.venv/bin/python -m pytest -q

Конфигурация находится в example.env; каждая переменная описанный там. Главные из них:

Переменная

По умолчанию

Почему это важно

MAX_SEARCHES_PER_TOOL_CALL

30

Ограничение на количество поисковых запросов за один вызов. Ограничено жёстким максимумом в 60.

MAX_CONCURRENT_SEARCHES

10

Степень параллелизма при массовой рассылке запросов.

MAX_HTTP_CONNECTIONS

60

Верхний предел пула соединений; бессерверные инстансы используют общий пул файловых дескрипторов.

REQUEST_TIMEOUT_SECONDS

105

Совпадает с верхним пределом на стороне вышестоящего провайдера, поэтому эта сторона никогда не превышает таймаут первой.

DEFAULT_RESULT_LIMIT

10

Количество результатов, запрашиваемых для каждого отдельного поиска на стороне вышестоящего провайдера.

SIGNUP_URL

RapidAPI listing

Сообщается пользователям, которые приходят без ключа.

MCP_PUBLIC_URL

http://localhost:8000/mcp

Указывается в /health.

RAPIDAPI_KEY

(empty)

Оставьте пустым в production. Если значение задано, каждый вызов без ключа обслуживается и оплачивается в счёт этой подписки. Сервер логирует предупреждение при запуске, а /health сообщается server_side_key_configured.

METRICS_TOKEN

(empty)

Когда задан, /metrics требует заголоок x-metrics-token.

LOG_PATH

(empty)

Пустое значение откличает файловый выход; строки MCP_CALL в stdout остаются записью. Корректно для serverless.

Операционные маршруты: GET /health (публичный, без аутентификации — реестры опрашивают его), GET /metrics, GET /metrics/calls?hours=24.

Целевая платформа развертывания — Vercel через api/index.py (FastAPI-обертка, передающая FastMCP его время жизни (lifespan), stateless_http=True). Канонический путь MCP — /mcp, без завершающего слеша.

Никогда не сохраняйте в репозитории настоящий ключ. example.env поставляется с заполнителями; сохраните это положение.

Отсутствие аффилированности

Это независимый API, который возвращает общедоступные цены на авиабилеты. Он не является аффилированным с Google, не одобрен и не спонсируется Google. "Google Flights" используется только для описания общедоступного источника данных. Тарифы предоставляются вышестоящим провайдером, постоянно меняются и не гарантируются — всегда подтверждайте цену на сайте авиакомпании или сервиса бронирования перед поокупкой.

Available Tools

4 tools
find_hotel_by_nameFlightPowers: find one hotel by nameA
Read-only
Inspect

FlightPowers single-property lookup: live Booking.com availability and pricing for one named property. Input: the hotel name a person would type (adding the city helps when a chain has many properties) plus check-in and check-out dates -- no internal property ID needed, the resolution is done for you. Returns the property's price, review score, room type and a booking link. Use it to check one specific hotel, or to track a single property's price over time.

price_as_seen_from prices the stay as a shopper resident in that country would see it. Gaps are real but usually modest and property-dependent, and rates move between calls, so call each country a few times on this same property before reporting a gap.

Rates go stale within minutes: never reuse an earlier result.

Requires the caller's own RapidAPI key for the Booking Live API. Get one (free tier available) at https://rapidapi.com/mtnrabi/api/booking-live-api, then pass it as an x-rapidapi-key header (preferred), a ?rapidapi_key= query parameter on the server URL, or your client's own API key field -- first non-empty wins. Usage counts against the caller's own RapidAPI plan, not ours; every response reports what it spent and what is left in api_usage.

ParametersJSON Schema
NameRequiredDescriptionDefault
adultsNoNumber of adult guests.
childrenNoNumber of children sharing the room.
currencyNoISO currency code for the prices returned, e.g. "usd".
hotel_nameYesThe property name a person would type, e.g. "Hotel Artemide". Adding the city ("Hotel Artemide Rome") disambiguates a chain with many properties. No internal property ID is needed.
checkin_dateYesFirst night of the stay, "YYYY-MM-DD".
checkout_dateYesDeparture morning, "YYYY-MM-DD". Must be after checkin_date.
price_as_seen_fromNoTwo-letter country code, e.g. "de". Prices the stay as a shopper resident in that country would see it. Call each country a few times on this same property before reporting a gap, because rates move between calls and gaps are usually modest and property-dependent.

Output Schema

ParametersJSON Schema
NameRequiredDescription
messageNo
resultsYesThe itineraries or properties found, already sorted and deduplicated. An empty array is only meaningful when search_status is 'empty'.
api_usageNoWhat this call cost the caller's own RapidAPI plan, and what remains on it. Present on every response that reached the upstream, including a degraded one -- a search that failed was still billed.
signup_urlNoWhere the caller subscribes or changes plan.
result_countNo
needs_api_keyNoTrue when no usable RapidAPI key arrived with the call, or the upstream rejected the one that did. No search was run and nothing was billed; signup_url and message say how to fix it.
applied_filtersNoWhich of the requested filters the upstream actually applied. Untyped: the shape is the upstream's, echoed through.
quota_exhaustedNoTrue when the caller's RapidAPI plan has no requests left for the current period.

TDQS

A4.3/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The description fully discloses live external API behavior, the need for an API key, usage costs, and that rates can change between calls. It also warns 'never reuse an earlier result,' which is meaningful behavioral context beyond the readOnly/Idempotent annotations.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is organized into clear, purposeful sections: the core lookup behavior, the price_as_seen_from caveat, freshness warning, and API key instructions. While a bit long, every section adds operational value and the structure makes it easy to scan.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness4/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

It covers what the tool returns, authentication requirements, cost attribution, freshness expectations, and a specific pricing-locale caveat. That is sufficient for correct invocation, though the exact output schema shape is not spelled out in the description.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters3/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The input schema already has 100% coverage with detailed parameter descriptions, including disambiguation guidance and checkout-after-checkin constraints. The prose mostly repeats these details rather than adding new parameter-level meaning, so it meets the baseline but does not go beyond it.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description opens with 'FlightPowers single-property lookup' and says it returns availability, pricing, review score, room type, and a booking link. This clearly defines the tool's exact function and distinguishes it from broad hotel search or flight search siblings.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines4/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description explicitly says to use it 'to check one specific hotel, or to track a single property's price over time,' which gives clear usage guidance. It does not explicitly mention the sibling search_hotels tool as the alternative for broad searches, but the 'single-property' framing makes that distinction clear.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

search_hotelsFlightPowers: search hotelsA
Read-only
Inspect

FlightPowers hotel search: live Booking.com availability and nightly prices for a destination and date range. Input: a free-text destination the way a person would say it ("Rome", "Tokyo Shibuya"), plus check-in and check-out dates. Returns each property's price, review score, room type, location and a booking link.

Set price_as_seen_from to a two-letter country code to price the same stay the way a shopper resident in that country would see it, which no other travel tool here can do. Gaps are real but usually modest and property-dependent, and rates move between calls, so hold one named property fixed, call each country a few times, and never read one call per country as a gap.

Rates go stale within minutes: never reuse an earlier result, search again.

Requires the caller's own RapidAPI key for the Booking Live API. Get one (free tier available) at https://rapidapi.com/mtnrabi/api/booking-live-api, then pass it as an x-rapidapi-key header (preferred), a ?rapidapi_key= query parameter on the server URL, or your client's own API key field -- first non-empty wins. Usage counts against the caller's own RapidAPI plan, not ours; every response reports what it spent and what is left in api_usage.

ParametersJSON Schema
NameRequiredDescriptionDefault
adultsNoNumber of adult guests. Defaults to the upstream default when omitted.
filtersNoProperty filters to apply, e.g. ["free_cancellation", "breakfast_included"]. An unknown name is rejected with the list of valid ones rather than being ignored.
childrenNoNumber of children sharing the room.
currencyNoISO currency code for the prices returned, e.g. "usd".
destinationYesWhere to stay, in free text the way a person would say it, e.g. "Rome" or "Tokyo Shibuya". A city, district, landmark or region all work; no internal location ID is needed.
checkin_dateYesFirst night of the stay, "YYYY-MM-DD".
checkout_dateYesDeparture morning, "YYYY-MM-DD". Must be after checkin_date.
budget_per_nightNoOnly return properties at or below this nightly price, in `currency`.
price_as_seen_fromNoTwo-letter country code, e.g. "de". Prices the stay through a residential connection in that country, so the result is what a shopper resident there would be quoted. For a rate-parity check hold one named property fixed and call each country a few times, because rates move between calls and one call per country can show a gap that is not there. Omit it for a neutral price.

Output Schema

ParametersJSON Schema
NameRequiredDescription
messageNo
resultsYesThe itineraries or properties found, already sorted and deduplicated. An empty array is only meaningful when search_status is 'empty'.
api_usageNoWhat this call cost the caller's own RapidAPI plan, and what remains on it. Present on every response that reached the upstream, including a degraded one -- a search that failed was still billed.
signup_urlNoWhere the caller subscribes or changes plan.
result_countNo
needs_api_keyNoTrue when no usable RapidAPI key arrived with the call, or the upstream rejected the one that did. No search was run and nothing was billed; signup_url and message say how to fix it.
applied_filtersNoWhich of the requested filters the upstream actually applied. Untyped: the shape is the upstream's, echoed through.
quota_exhaustedNoTrue when the caller's RapidAPI plan has no requests left for the current period.

TDQS

A4.9/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Beyond the readOnlyHint annotation, the description discloses several important behaviors: results are live and go stale within minutes, unknown filters are rejected rather than ignored, API usage is charged to the caller's own plan, and the first non-empty API key source wins. This goes well beyond what the annotations alone convey.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is somewhat long, but nearly every sentence carries necessary operational or behavioral information. The repetition around rate-parity checks is slightly verbose but serves an important warning purpose, so the length is justified overall.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

The description covers what the tool returns, how to supply inputs, how pricing and filtering behave, how rate-parity checks should be performed, and how API keys and usage accounting work. Combined with the output schema, an agent has everything needed to invoke and interpret this tool correctly.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

The schema already covers 100% of parameters with meaningful descriptions, and the tool description adds further value by explaining the semantics of price_as_seen_from, including the country-code behavior and how to interpret rate-parity results. Parameter meaning is fully clear.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description clearly states that this is a hotel search tool for live Booking.com availability and nightly prices by destination and date range. It is immediately distinguishable from the flight-search siblings, and the title reinforces the purpose.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description gives direct guidance on when to use the tool and how to use it correctly, including the rate-parity workflow with price_as_seen_from, the need to hold one property fixed, and the warning that rates move between calls. It also explains API key handling and billing, leaving no ambiguity about operational usage.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

search_oneway_flightsFlightPowers: search one-way flightsA
Read-only
Inspect

FlightPowers one-way fare search: live prices read from Google Flights. Input: origin and destination IATA codes -- the destination may be several codes, as "BCN,LIS,ATH" or ["BCN","LIS","ATH"] -- plus either one departure date or a date range. Returns each flight's price, airline, duration, stops, a bookable buy_link, and Google's historical price range (price_insights_low / price_insights_high) so you can say whether a fare is actually a good deal.

Use it for any one-way fare question, including open-ended ones. For a flexible search make ONE call with a date range and/or several destinations -- do NOT call it once per date. 'Cheapest flight to Sri Lanka anywhere in October' is one call, not thirty.

Each date/destination combination is one billed request; the count and the plan's remaining quota come back in api_usage.

by_destination carries one entry per destination you asked for -- empty ones included, each with a reason -- so read it before telling a user a destination has no flights.

Requires the caller's own RapidAPI key for the Google Flights Live API. Get one (free tier available) at https://rapidapi.com/mtnrabi/api/google-flights-live-api, then pass it as an x-rapidapi-key header (preferred), a ?rapidapi_key= query parameter on the server URL, or your client's own API key field -- first non-empty wins. Usage counts against the caller's own RapidAPI plan, not ours; every response reports what it spent and what is left in api_usage.

ParametersJSON Schema
NameRequiredDescriptionDefault
limitNoMaximum flights to return, after merging and sorting.
sort_byNo"best", "price", or "duration". Applied across all results.best
currencyNoISO currency code, default "usd".usd
max_priceNoOnly return flights at or below this price.
max_stopsNoMaximum stops per flight. 0 means non-stop only.
seat_typeNo1 economy, 2 premium economy, 3 business, 4 first.
passengersNoPassenger counts as [adults, children, infants].
to_airportYesDestination airport. One IATA code ("BCN"), several separated by commas ("BCN,LIS,ATH"), or a list (["BCN","LIS","ATH"]) -- every shape is accepted and the destinations are compared in the same search.
from_airportYesOrigin IATA code, e.g. "TLV". One origin per search; a second one is refused rather than searched.
max_searchesNoCap the billed requests this call may make. Lower it to spend less of the plan's quota on a wide search; the range is then sampled evenly rather than cut short.
use_fallbackNoLeave unset. Switches the search to a second, independent flight data source instead of the usual Google Flights page read. Unset already escalates to that source once, automatically, after a search's retries have failed. true forces it inline on every attempt -- much slower, and it can time out. false disables it entirely, that automatic retry included.
airline_codesNoRestrict to these airline codes, e.g. ["LY"].
departure_dateNoSingle departure date, "YYYY-MM-DD".
arrival_time_maxNoLatest arrival hour, 0-23.
arrival_time_minNoEarliest arrival hour, 0-23.
departure_date_toNoLast date of a departure range.
departure_time_maxNoLatest departure hour, 0-23.
departure_time_minNoEarliest departure hour, 0-23.
departure_date_fromNoFirst date of a departure range.
exclude_airline_codesNoExclude these airline codes.

Output Schema

ParametersJSON Schema
NameRequiredDescription
messageNoPresent when there is something the model must relay to the user rather than silently absorb -- no results, a degraded search, a missing key or a spent quota.
partialNoPresent when some searches failed but others succeeded. Plain text saying how much of the request the results cover.
resultsYesThe itineraries or properties found, already sorted and deduplicated. An empty array is only meaningful when search_status is 'empty'.
api_usageNoWhat this call cost the caller's own RapidAPI plan, and what remains on it. Present on every response that reached the upstream, including a degraded one -- a search that failed was still billed.
signup_urlNoWhere the caller subscribes or changes plan.
result_countNo
needs_api_keyNoTrue when no usable RapidAPI key arrived with the call, or the upstream rejected the one that did. No search was run and nothing was billed; signup_url and message say how to fix it.
search_statusNoWhether the underlying search actually completed, read from the backend's X-Search-Status header. 'ok': every combination searched returned results. 'empty': the search completed and Google genuinely has no itineraries for it -- a real answer, not a failure. 'partial': some combinations returned results and some failed, so the list is incomplete. 'degraded': every combination failed, so the search did not happen and an empty list means nothing; this case is also flagged with isError: true and is safe to retry.
by_destinationNoOne entry per destination the REQUEST asked for, in request order, present whether or not that destination has any flights in `results`. A destination with an empty `rows` array is a hole in the answer, and `reason` says which kind of hole: 'no_flights' (searched, answered, Google has nothing), 'search_failed' (searched and the search errored, so nothing is known), 'not_in_limit' (searched, found flights, none fitted in `limit`) or 'not_searched' (never searched -- the per-call fan-out cap sampled it away). 'ok' means it has rows. Read this rather than inferring coverage from `results`: a destination missing from `results` looks identical to one that has no flights, and they are not the same answer. `rows` are the same row objects that are in `results`, in the same order -- nothing here is data the answer does not already contain.
quota_exhaustedNoTrue when the caller's RapidAPI plan has no requests left for the current period.
search_coverageNoWhat was actually searched. Present whether or not the request was truncated, so a model can state honestly what its answer rests on.

TDQS

A5/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

The description discloses return fields (price, airline, duration, stops, buy_link, price_insights) and billing reporting via api_usage. It further explains fallback source behavior, max_searches sampling, and that usage counts against the caller's own RapidAPI plan, all beyond the annotations.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness5/5

Is the description appropriately sized, front-loaded, and free of redundancy?

Although lengthy, each paragraph serves a distinct purpose—result contents, invocation advice, billing, and API-key setup—and includes concrete examples without redundant filler. The structure is logical and easy to scan.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

The description covers key operational concerns an external agent needs: API key requirements, quota reporting, multi-destination/date handling, and fallback behavior. Since an output schema is present, no additional return-value documentation is necessary.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

All 20 parameters have schema descriptions, and the description adds practical semantics such as accepted shapes for to_airport, origin singularity, date-range versus single-date usage, and seat_type/passenger mappings. This goes well beyond the schema's basic property descriptions.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Title and first sentence explicitly state 'one-way fare search' and 'live prices read from Google Flights,' clearly distinguishing it from sibling roundtrip/hotel tools. The verb 'search' plus resource 'flights' makes the purpose unambiguous.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

The description gives explicit guidance: 'Use it for any one-way fare question' and 'do NOT call it once per date,' with a concrete example of a single flexible search. It also explains API-key requirements, billing consequences, and that a second origin is refused rather than searched.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

search_roundtrip_flightsFlightPowers: search round-trip flightsA
Read-only
Inspect

FlightPowers round-trip fare search: live prices read from Google Flights, priced as paired legs rather than two separate one-ways. Input: origin and destination IATA codes -- the destination may be several codes, as "BCN,LIS,ATH" or ["BCN","LIS","ATH"] -- a departure date or range, and either a return date or a trip length in nights. Returns the total price for both legs, per-leg airline, stops and duration, and a single bookable buy_link for the trip.

Use it for any return-trip fare question. For a flexible search make ONE call: pass departure_date_from / departure_date_to for the outbound range and nights instead of return_date to compare trip lengths -- '5 to 7 nights in Rome sometime in May' is one call.

Each date/destination combination is one billed request; the count and the plan's remaining quota come back in api_usage.

by_destination carries one entry per destination you asked for -- empty ones included, each with a reason -- so read it before telling a user a destination has no flights.

Requires the caller's own RapidAPI key for the Google Flights Live API. Get one (free tier available) at https://rapidapi.com/mtnrabi/api/google-flights-live-api, then pass it as an x-rapidapi-key header (preferred), a ?rapidapi_key= query parameter on the server URL, or your client's own API key field -- first non-empty wins. Usage counts against the caller's own RapidAPI plan, not ours; every response reports what it spent and what is left in api_usage.

ParametersJSON Schema
NameRequiredDescriptionDefault
limitNoMaximum trips to return, after merging and sorting.
nightsNoTrip length in nights; a number, or a list like [5, 6, 7]. The return date is derived from each departure date.
sort_byNo"best", "price", or "duration". Applied across all results.best
currencyNoISO currency code, default "usd".usd
max_priceNoOnly return trips at or below this total price.
seat_typeNo1 economy, 2 premium economy, 3 business, 4 first.
passengersNoPassenger counts as [adults, children, infants].
to_airportYesDestination airport. One IATA code ("BCN"), several separated by commas ("BCN,LIS,ATH"), or a list (["BCN","LIS","ATH"]) -- every shape is accepted and the destinations are compared in the same search.
return_dateNoFixed return date. Use this OR nights, not both.
from_airportYesOrigin IATA code, e.g. "TLV". One origin per search; a second one is refused rather than searched.
max_searchesNoCap the billed requests this call may make. Lower it to spend less of the plan's quota on a wide search; the range is then sampled evenly rather than cut short.
use_fallbackNoLeave unset. Switches the search to a second, independent flight data source instead of the usual Google Flights page read. Unset already escalates to that source once, automatically, after a search's retries have failed. true forces it inline on every attempt -- much slower, and it can time out. false disables it entirely, that automatic retry included.
departure_dateNoSingle outbound date, "YYYY-MM-DD".
max_return_stopsNoMaximum stops on the return leg.
departure_date_toNoLast date of an outbound range.
departure_date_fromNoFirst date of an outbound range.
max_departure_stopsNoMaximum stops on the outbound leg.
return_airline_codesNoRestrict the return leg to these airlines.
departure_airline_codesNoRestrict the outbound leg to these airlines.

Output Schema

ParametersJSON Schema
NameRequiredDescription
messageNoPresent when there is something the model must relay to the user rather than silently absorb -- no results, a degraded search, a missing key or a spent quota.
partialNoPresent when some searches failed but others succeeded. Plain text saying how much of the request the results cover.
resultsYesThe itineraries or properties found, already sorted and deduplicated. An empty array is only meaningful when search_status is 'empty'.
api_usageNoWhat this call cost the caller's own RapidAPI plan, and what remains on it. Present on every response that reached the upstream, including a degraded one -- a search that failed was still billed.
signup_urlNoWhere the caller subscribes or changes plan.
result_countNo
needs_api_keyNoTrue when no usable RapidAPI key arrived with the call, or the upstream rejected the one that did. No search was run and nothing was billed; signup_url and message say how to fix it.
search_statusNoWhether the underlying search actually completed, read from the backend's X-Search-Status header. 'ok': every combination searched returned results. 'empty': the search completed and Google genuinely has no itineraries for it -- a real answer, not a failure. 'partial': some combinations returned results and some failed, so the list is incomplete. 'degraded': every combination failed, so the search did not happen and an empty list means nothing; this case is also flagged with isError: true and is safe to retry.
by_destinationNoOne entry per destination the REQUEST asked for, in request order, present whether or not that destination has any flights in `results`. A destination with an empty `rows` array is a hole in the answer, and `reason` says which kind of hole: 'no_flights' (searched, answered, Google has nothing), 'search_failed' (searched and the search errored, so nothing is known), 'not_in_limit' (searched, found flights, none fitted in `limit`) or 'not_searched' (never searched -- the per-call fan-out cap sampled it away). 'ok' means it has rows. Read this rather than inferring coverage from `results`: a destination missing from `results` looks identical to one that has no flights, and they are not the same answer. `rows` are the same row objects that are in `results`, in the same order -- nothing here is data the answer does not already contain.
quota_exhaustedNoTrue when the caller's RapidAPI plan has no requests left for the current period.
search_coverageNoWhat was actually searched. Present whether or not the request was truncated, so a model can state honestly what its answer rests on.

TDQS

A4.9/5.0
Behavior5/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Discloses external RapidAPI dependency, caller-provided key requirements, billing/quota consumption, retry and fallback behavior, and the fact that each destination/date combination is a billed request. This goes beyond the read-only annotation and gives accurate expectations of side effects.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness4/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The description is thorough but somewhat long, with some details repeated across paragraphs and parameter descriptions. However, the content is organized into clear topical paragraphs and nearly every sentence carries practical guidance, making the length justified overall.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness5/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

Covers external API integration, authentication, billing, fallback source switching, result composition, and the meaning of api_usage and by_destination. Given the tool's complexity and external dependencies, the description contains all necessary context for correct invocation and interpretation.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters5/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Every schema parameter has a meaningful description, and the description adds important semantic details such as single-origin refusal, accepted destination formats, nights vs return_date exclusivity, and even sampling behavior for max_searches. The prose supplements the schema rather than merely repeating it.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose5/5

Does the description clearly state what the tool does and how it differs from similar tools?

Clearly identifies the tool as a round-trip flight fare search reading live prices from Google Flights and pricing paired legs, not separate one-ways. This distinguishes it from the sibling one-way search tool without ambiguity.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines5/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Explicitly directs use for any return-trip fare question and gives concrete guidance for flexible date ranges, multi-destination searches, and quota control via max_searches. The instructions about return_date vs nights and fallback behavior leave little room for misuse.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

  1. 4 tool updatesv1.0.0
    • First observedfind_hotel_by_name
    • First observedsearch_hotels
    • First observedsearch_oneway_flights
    • First observedsearch_roundtrip_flights
Deploy Server
One endpoint for every MCP toolRayrun runs every MCP server behind one URL and holds the credentials. Each call is decided per client, tool, and argument, then recorded.ray.runAd

TDQS

A4.6/5.0

Scored across 4 tools

Disambiguation5/5

Each tool targets a clearly distinct query type: one-way flights, round-trip flights, general hotel search, and specific hotel lookup. Even the two hotel tools are easy to differentiate because one is broad destination search and the other is single-property lookup.

Naming Consistency4/5

Three tools follow the search_<subject> pattern: search_oneway_flights, search_roundtrip_flights, and search_hotels. find_hotel_by_name breaks the pattern by using find_ instead of search_, though it remains readable and predictable.

Tool Count5/5

Four tools is a well-scoped set for a travel-search server. Each tool earns its place and the count avoids both bloat and thinness.

Completeness4/5

The core search workflows are covered: one-way flights, round-trip flights, hotel search, and named-hotel lookup. Obvious gaps include multi-city flight search and airport/place code resolution, but agents can work around these with external knowledge or by combining existing tools.

Maintenance

ActivityActive
ResponsivenessUnresponsive

Resources

Related MCP Connectors

Related MCP Servers

  • SkyOdyssey MCP

    maru-775
    A
    license
    Not graded
    quality
    D
    maintenance
    Enables AI clients to explore cheapest destinations, optimize multi-leg flight itineraries, and reference airport/region data via MCP tools and resources.
    MIT

  • Google Flights MCP

    mtnrabi
    A
    license
    Not graded
    quality
    D
    maintenance
    Enables real-time one-way and round-trip flight searches from any MCP client without an API key or account, returning live fares, historical price insights, and bookable links supported by disclosed sponsored cards.
    1
    MIT