Google Flights MCP
Enables searching Google Flights for one-way, round-trip, and multi-city itineraries, plus fare tracking and price alerts via SerpAPI.
README
Google Flights MCP
An MCP server that lets Claude search Google
Flights and track fares over time. Google has no official flights API, so this
server calls SerpAPI's google_flights
engine, which runs the search server-side and returns structured JSON.
Tools
| Tool | What it does |
|---|---|
search_flights |
Search between two airports (one-way or round-trip), with cabin class, passengers, stops, and price filters. |
search_multi_city |
Search a multi-city itinerary (2+ legs). Returns leg-1 options; advance leg by leg with get_return_flights. |
get_return_flights |
Select a flight via its departure_token to see the next set of options (round-trip return, or next multi-city leg). |
get_booking_options |
For a chosen flight (via its booking_token), list where to book and the price at each provider. |
track_price |
Save a search as a price watch, with an optional target price and label. |
list_price_watches |
List all watches and their most recent recorded price. |
check_prices |
Re-check every watch now, record prices, and report drops / target hits. |
get_price_history |
Show the recorded price history for a watch. |
untrack_price |
Delete a watch and its history. |
Cabin class
search_flights, search_multi_city, and track_price all accept
travel_class: economy (default), premium_economy, business,
first. Just ask, e.g. "business class SEA to London."
Setup
-
Get a SerpAPI key (free tier: 100 searches/month) at https://serpapi.com/manage-api-key.
-
Add the key:
cp .env.example .env # then edit .env and set SERPAPI_API_KEY=... -
Install dependencies (uv provisions an isolated Python ≥3.10):
uv sync -
Register with Claude Code (user scope = available in every session):
claude mcp add google-flights -s user -- \ uv run --directory /Users/pisan/bitbucket/pisanuw/google-flights python server.pyRestart Claude Code (or run
/mcp) to pick up the tools.
Example prompts
- "Nonstop round-trip SEA→NRT, Oct 3 back Oct 17, under $1200."
- "Business class one-way SEA to LHR on Nov 15."
- "Multi-city: SEA→NRT Oct 3, NRT→ICN Oct 10, ICN→SEA Oct 17."
- "Track SEA→LAX Nov 15–20 and alert me if it drops below $120."
- "Check my flight watches" / "show the price history for watch 2."
Price alerts (background checking)
check_prices runs on demand (when you ask Claude). For hands-off alerts, run
the standalone checker on a schedule; it records prices and fires a native macOS
notification on any drop or target hit:
uv run python check_watches.py
Schedule it with launchd or cron (e.g. hourly). Watches and history are stored
in watches.db (gitignored) next to the server, so the scheduled checker and
the MCP tools share the same data.
Notes
- Airports are IATA codes (
SEA,NRT,LHR). Multiple origins can be comma-separated (SEA,PDX). - For round trips, each result's
priceis the total round-trip fare. - Multi-city is leg-by-leg: the first response is leg-1 options; the shown price
is the full-itinerary estimate. Use
get_return_flightswith adeparture_tokento advance. deep_search=truematches google.com/travel/flights more closely but is slower.- Every search (including each watch re-check) counts against your SerpAPI quota.
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。