skiplagged-travel-search

Compare original and translation side by side

🇺🇸

Original

English
🇨🇳

Translation

Chinese

Skiplagged Travel Search

Skiplagged 旅行搜索

Overview

概述

Skiplagged MCP exposes travel search capabilities (flights, flexible date calendars, “anywhere” destination discovery, hotels + room details, and sometimes rental cars) as MCP tools. Prefer calling Skiplagged tools over guessing prices/availability.
Skiplagged MCP 将旅行搜索功能(航班、灵活日期日历、“任意目的地”探索、酒店及房型详情,有时还包括租车)以MCP工具的形式开放。 优先调用Skiplagged工具,而非猜测价格或可预订情况。

Prerequisites

前提条件

  1. Connect the MCP client to
    https://mcp.skiplagged.com/mcp
    over HTTP transport.
  2. List tools and confirm Skiplagged tools are visible before running workflows.
  3. Treat authentication as not required.
  4. Read
    references/setup.md
    for client-specific setup commands and examples.

  1. 通过HTTP传输将MCP客户端连接至
    https://mcp.skiplagged.com/mcp
  2. 在运行工作流前,先列出工具并确认Skiplagged工具可见。
  3. 无需进行身份验证。
  4. 请阅读
    references/setup.md
    获取针对特定客户端的设置命令及示例。

Operating rules

操作规则

1) Grounding rule

1) 落地规则

Always base answers on tool results. Never invent or guess prices, availability, or policies.
始终基于工具返回结果作答。绝不可编造或猜测价格、可预订情况及相关政策。

2) Tool-call discipline

2) 工具调用规范

  • Use minimal, correct parameters (avoid unnecessary filters unless user asked).
  • Prefer IATA airport codes when possible; otherwise accept city names and clarify if needed.
  • Use small result limits when supported (e.g., 3–7) and expand only if the user asks.
  • Chain multiple tool calls when needed, but avoid loops. Respect per-turn tool-call limits.
  • 使用最少且正确的参数(除非用户要求,否则避免不必要的筛选条件)。
  • 尽可能优先使用IATA机场代码;若用户提供的是城市名称,必要时需进行确认。
  • 若工具支持,使用较小的结果限制数量(如3-7条),仅在用户要求时再扩大范围。
  • 必要时可串联多次工具调用,但避免循环调用。遵守每轮对话的工具调用次数限制。

3) Post-call validation loop

3) 调用后验证流程

After each tool call:
  1. Summarize what the tool returned in 1–2 sentences (what matters to a traveler: cost, time, stops, key tradeoffs).
  2. Decide the next step:
    • proceed (narrow or enrich),
    • self-correct (relax filters, adjust airports/dates),
    • or ask one targeted follow-up question.
每次工具调用后:
  1. 用1-2句话总结工具返回的内容(重点关注旅行者关心的信息:价格、时长、经停次数、核心取舍点)。
  2. 决定下一步操作:
    • 继续(缩小范围或丰富信息),
    • 自我修正(放宽筛选条件、调整机场/日期),
    • 或提出一个针对性的跟进问题。

4) Failure handling (empty or unhelpful results)

4) 异常处理(结果为空或无帮助)

If results are empty/ambiguous:
  • Say so clearly.
  • Suggest the smartest next step (broader dates, nearby airports, allow 1 stop, adjust time window, try “anywhere” discovery, or switch to hotels-first for trip planning).
  • For failures other than lack of results, suggest using https://skiplagged.com directly.
若结果为空或模糊不清:
  • 清晰告知用户这一情况。
  • 建议最合理的下一步操作(扩大日期范围、选择邻近机场、允许1次经停、调整时间窗口、尝试“任意目的地”探索,或切换为以酒店为核心的行程规划)。
  • 若因结果为空以外的原因导致失败,建议用户直接访问https://skiplagged.com。

5) Output style (traveler-centric)

5) 输出风格(以旅行者为中心)

Be concise and practical:
  • Highlight savings, routing insights, and key tradeoffs (stops vs duration vs price).
  • Present a short shortlist (typically 3–7 options) rather than a dump of results.

内容需简洁实用:
  • 突出优惠力度航线洞察核心取舍点(经停次数 vs 行程时长 vs 价格)。
  • 展示简短的精选列表(通常为3-7个选项),而非全部结果。

Tool discovery (required, safer across deployments)

工具发现(必填,跨部署场景更安全)

Different deployments may rename tools or add new ones. Before running any workflow:
  1. List tools available on the connected Skiplagged MCP server.
  2. Identify the best match for each capability:
    • flights search
    • flexible date calendars
    • anywhere destinations
    • hotels search
    • hotel details
    • (optional) rental cars
Mapping convention (typical, not guaranteed): tools often start with
sk_
.
不同部署场景下,工具名称可能会被重命名或新增工具。在运行任何工作流前:
  1. 列出已连接的Skiplagged MCP服务器上的所有工具。
  2. 为每项功能匹配最合适的工具:
    • 航班搜索
    • 灵活日期日历
    • 任意目的地探索
    • 酒店搜索
    • 酒店详情
    • (可选)租车
映射惯例(常见情况,不保证绝对适用):工具通常以
sk_
开头。

Common capability → typical tool names (use as a guide)

常见功能→典型工具名称(仅供参考)

  • Flights search →
    sk_flights_search
  • Flexible departure calendar →
    sk_flex_departure_calendar
  • Flexible return calendar →
    sk_flex_return_calendar
  • Anywhere destination discovery →
    sk_destinations_anywhere
  • Hotels search →
    sk_hotels_search
  • Hotel details →
    sk_hotel_details
If your tool list differs, use the closest equivalent and keep the workflows below unchanged.

  • 航班搜索 →
    sk_flights_search
  • 灵活出发日期日历 →
    sk_flex_departure_calendar
  • 灵活返程日期日历 →
    sk_flex_return_calendar
  • 任意目的地探索 →
    sk_destinations_anywhere
  • 酒店搜索 →
    sk_hotels_search
  • 酒店详情 →
    sk_hotel_details
若你的工具列表与上述不同,请选择最接近的等效工具,后续工作流无需更改。

Intake checklist (ask only what’s needed)

需求收集清单(仅询问必要信息)

Extract or confirm only the missing critical inputs:
  • Route intent: specific route (
    origin -> destination
    ) vs flexible destination (“anywhere”).
  • Dates: exact dates or flexible window + trip length.
  • Travelers: adults/children/infants and cabin, if non-default.
  • Constraints: budget, stops, max duration, time windows, and airline preferences.
  • Lodging intent (if relevant): hotel location, dates, guests, and required amenities.
仅提取或确认缺失的关键信息:
  • 航线意向:特定航线(出发地→目的地)还是灵活目的地(“任意地点”)。
  • 日期:具体日期或灵活时间范围+行程时长。
  • 旅客:成人/儿童/婴儿人数及舱位等级(若未指定则使用默认值)。
  • 限制条件:预算、经停次数、最长行程时长、时间窗口、航空公司偏好。
  • 住宿意向(若相关):酒店位置、日期、客人数量、必备设施。

Workflows

工作流

Workflow A — Standard flight search (one-way or round-trip)

工作流A — 标准航班搜索(单程或往返)

Use when the user wants specific flights between places.
Preferred tool → fallback
  • Preferred:
    sk_flights_search
Steps
  1. Normalize locations:
    • If user gives a city/region, clarify which airport(s) are acceptable if it changes results materially (multi-airport metros).
  2. Call the flight-search tool with:
    • origin, destination
    • depart date (+ return date if round-trip)
    • passenger counts, cabin - assume defaults if not specified
    • only the filters the user asked for (stops, time windows, airline preferences)
  3. Post-process:
    • Cluster top results by cheapest, fastest, fewest stops.
    • Present 3–7 options with price, stops, duration, and any standout caveats (overnight layover, long layover).
  4. If results are too broad:
    • Ask 1 targeted question (e.g., “Allow one stop?”) or apply a reasonable default and disclose it.

适用于用户需要特定地点间航班的场景。
首选工具→备选方案
  • 首选:
    sk_flights_search
步骤
  1. 标准化地点信息:
    • 若用户提供的是城市/地区,若会对结果产生重大影响(多机场都市圈),需确认可接受的机场。
  2. 调用航班搜索工具,传入以下参数:
    • 出发地、目的地
    • 出发日期(若为往返则需加上返程日期)
    • 旅客人数、舱位等级(未指定则使用默认值)
    • 仅添加用户要求的筛选条件(经停次数、时间窗口、航空公司偏好)
  3. 后处理:
    • 最便宜最快经停次数最少对顶级结果进行分组。
    • 展示3-7个选项,包含价格、经停次数、行程时长及任何突出的注意事项(隔夜中转、长时间中转)。
  4. 若结果范围过广:
    • 提出1个针对性问题(如“是否允许1次经停?”),或应用合理的默认值并告知用户。

Workflow B — Flexible dates / fare calendar

工作流B — 灵活日期/票价日历

Use when the user says “around X date”, “flexible”, or “cheapest days”.
Preferred tools → fallback
  • Preferred:
    sk_flex_departure_calendar
    and
    sk_flex_return_calendar
Steps
  1. Call the departure-calendar tool around the target date/window.
  2. If round-trip:
    • Use a trip length (or return window) that matches the user’s intent (e.g., “3–5 nights”).
    • Call the return-calendar tool for best return dates.
  3. Summarize:
    • Cheapest departure day(s)
    • Cheapest round-trip date pairs (if supported)
    • Recommend 2–4 date options aligned to constraints (weekends, time windows, budget).

适用于用户提及“X日期前后”“灵活时间”或“最便宜日期”的场景。
首选工具→备选方案
  • 首选:
    sk_flex_departure_calendar
    sk_flex_return_calendar
步骤
  1. 围绕目标日期/时间窗口调用出发日期日历工具。
  2. 若为往返行程:
    • 使用符合用户意向的行程时长(或返程时间范围)(如“3-5晚”)。
    • 调用返程日期日历工具获取最佳返程日期。
  3. 总结:
    • 最便宜的出发日期
    • 最便宜的往返日期组合(若工具支持)
    • 根据限制条件(周末、时间窗口、预算)推荐2-4个日期选项。

Workflow C — “Anywhere” destination discovery

工作流C — “任意目的地”探索

Use when the user is flexible on where to go (“somewhere warm”, “anywhere cheap”).
Preferred tool → fallback
  • Preferred:
    sk_destinations_anywhere
Steps
  1. Confirm:
    • origin
    • date/window (or month)
    • constraints (budget, max flight time, nonstop-only, region/climate)
  2. Call the anywhere-destinations tool.
  3. Return 5–12 destinations with:
    • destination, lowest fare, suggested dates/window (as returned)
    • a one-line rationale tied to the user constraints
  4. When the user picks a destination, transition to Workflow A to find bookable itineraries.

适用于用户对目的地无明确要求的场景(如“温暖的地方”“便宜的地方”)。
首选工具→备选方案
  • 首选:
    sk_destinations_anywhere
步骤
  1. 确认以下信息:
    • 出发地
    • 日期/时间窗口(或月份)
    • 限制条件(预算、最长飞行时间、直飞优先、地区/气候)
  2. 调用任意目的地探索工具。
  3. 返回5-12个目的地,包含:
    • 目的地、最低票价、建议日期/时间窗口(按工具返回内容)
    • 结合用户限制条件的一句话推荐理由
  4. 当用户选定目的地后,切换至工作流A查找可预订的行程。

Workflow D — Hotels + room-level details

工作流D — 酒店及房型详情

Use when the user asks for hotels, places to stay, or wants to bundle planning.
Preferred tools → fallback
  • Preferred:
    sk_hotels_search
    then
    sk_hotel_details
  • Fallback: any tool(s) that provide hotel lists and separate room/policy breakdowns
Steps
  1. Call the hotel-search tool with:
    • location, dates, rooms/guests
    • only requested filters (price cap, rating, amenities)
  2. Present 5–10 options with:
    • total price / nightly price (as returned), rating, key amenities, and area/neighborhood (if returned)
  3. When the user selects a hotel (or asks for “best value”), call the hotel-details tool.
  4. Summarize room options:
    • room type, cancellation policy, inclusions (breakfast), total price (as returned)
    • highlight tradeoffs: nonrefundable vs flexible, fees, value differences.

适用于用户询问酒店、住宿地点,或需要打包规划行程的场景。
首选工具→备选方案
  • 首选:先调用
    sk_hotels_search
    ,再调用
    sk_hotel_details
  • 备选:任何可提供酒店列表及独立房型/政策详情的工具
步骤
  1. 调用酒店搜索工具,传入以下参数:
    • 位置、日期、房间数/客人数量
    • 仅添加用户要求的筛选条件(价格上限、评分、设施)
  2. 展示5-10个选项,包含:
    • 总价/每晚价格(按工具返回内容)、评分、核心设施、区域/街区(若工具返回)
  3. 当用户选定酒店(或询问“最具性价比选项”)时,调用酒店详情工具。
  4. 总结房型选项:
    • 房型、取消政策、包含服务(如早餐)、总价(按工具返回内容)
    • 突出取舍点:不可退款 vs 灵活取消、费用、性价比差异。

Workflow E — Rental cars (if available)

工作流E — 租车(若支持)

Use when the user requests a car rental.
Preferred tool → fallback
  • Preferred: a tool in the server tool list whose description indicates car-rental search
Steps
  1. Confirm pickup/dropoff location, times, and any constraints (car class, driver age if required).
  2. Call the car-rental search tool (if present).
  3. Present 3–7 options with total price, class, company, and relevant policies (as returned).

适用于用户要求租车的场景。
首选工具→备选方案
  • 首选:服务器工具列表中描述为租车搜索的工具
步骤
  1. 确认取车/还车地点、时间及任何限制条件(车型、驾驶员年龄要求)。
  2. 调用租车搜索工具(若存在)。
  3. 展示3-7个选项,包含总价、车型、公司及相关政策(按工具返回内容)。

Planning behavior (for full itineraries)

全行程规划行为

When the user asks to “plan a trip”:
  1. Anchor with flights (or anywhere discovery) to set destination/dates/cost.
  2. Add hotels for the chosen destination/dates.
  3. Add rental cars if needed and supported.
  4. Present 2–3 bundles (e.g., “cheapest”, “balanced”, “most convenient”) with clear tradeoffs.

当用户要求“规划行程”时:
  1. 以航班(或任意目的地探索)为基础,确定目的地、日期及成本。
  2. 为选定的目的地和日期添加酒店信息。
  3. 若需要且工具支持,添加租车信息。
  4. 展示2-3个打包方案(如“最便宜”“均衡型”“最便捷”),并明确说明各方案的取舍点。

Reliability + transparency notes

可靠性与透明度说明

  • Prices and availability change quickly; treat results as current at the moment of the tool call and encourage the user to confirm terms via the provided link (if any).
  • If hidden-city itineraries appear, be transparent about common constraints and tradeoffs (especially baggage and missed-leg implications). Do not oversell—present as an option with clear caveats.

  • 价格和可预订情况变化迅速;工具调用时的结果仅代表当前状态,建议用户通过提供的链接(若有)确认最终条款。
  • 若出现隐藏城市行程,需透明告知常见限制条件及取舍点(尤其是行李托运和错过航段的影响)。不得过度宣传,需将其作为一个选项并明确标注注意事项。

Troubleshooting (avoid loops)

故障排除(避免循环)

  • If a call fails or returns empty results:
    • Retry at most once with broader constraints (e.g., allow 1 stop, widen time window, nearby airports).
    • If still empty, ask one targeted follow-up question and propose a concrete alternative search.
  • 若调用失败或返回空结果:
    • 最多重试一次,同时放宽限制条件(如允许1次经停、扩大时间窗口、选择邻近机场)。
    • 若仍无结果,提出一个针对性的跟进问题,并建议一个具体的替代搜索方向。