reservations-mcp
Enables restaurant discovery and reservations across multiple providers (Resy, Google Places, Yelp, Tock) with auditable and secure two-step booking.
README
reservations-mcp
An auditable Model Context Protocol server for restaurant discovery and reservations, across multiple providers.
Built as a transparent alternative to the "just npx this mystery package" reservation
MCPs floating around: every line is here in src/, it talks only to each provider's
own API, it uses no browser automation or bot-detection evasion, and credentials come
only from your own environment or local config — never logged, never sent anywhere else.
What it does
| Provider | Discovery | Availability | Booking |
|---|---|---|---|
| Resy | ✅ venue search | ✅ | ✅ (two-step, opt-in) |
| Google Places | ✅ (rich: goodForChildren, outdoorSeating, servesCocktails) |
— | — |
| Yelp | ✅ (outdoor_seating, price, term) |
— | — |
| Tock | ⚠️ stub in v1 (returns a direct link; see src/providers/tock.ts) |
— | — |
OpenTable is intentionally excluded. It has no public reservation API, so booking it requires browser automation that violates its ToS and risks an account ban. Out of scope.
Tools
discover_restaurants— natural-language search ("Hudson waterfront, skyline view, outdoor, kid-friendly, great cocktails"). Read-only.resy_search— find Resy venue ids by name/keyword.resy_availability— open slots for a venue / date / party size.resy_booking_preview— step 1: resolve a slot, surface time + seating + any cancellation/no-show fee. Does not book.resy_book— step 2, irreversible: books abook_token. Requiresconfirm: true.resy_payment_methods— list saved card ids (some bookings require a card).resy_login— exchange email+password for a token once (password never stored). PreferRESY_AUTH_TOKEN.resy_my_reservations— your upcoming reservations.tock_search— best-effort stub (see note above).status— which providers are configured.
Setup
npm install
npm run build
cp .env.example .env # then fill in keys/tokens you want
Credentials (all optional — set only what you need)
RESY_AUTH_TOKEN— required for Resy availability/booking. Log in at resy.com, open DevTools → Network, click any restaurant, and copy theX-Resy-Auth-Tokenheader. (Or call theresy_logintool once.)RESY_PAYMENT_METHOD_ID— default card for bookings (resy_payment_methodslists ids).YELP_API_KEY— free from the Yelp Fusion console.GOOGLE_PLACES_API_KEY— Google Places API (New).
RESY_API_KEY defaults to the public key embedded in Resy's own web client; override if it rotates.
Connect to Claude
Add to your Claude Code / Claude Desktop MCP config:
{
"mcpServers": {
"reservations": {
"command": "node",
"args": ["/absolute/path/to/reservations-mcp/dist/index.js"],
"env": {
"RESY_AUTH_TOKEN": "…",
"GOOGLE_PLACES_API_KEY": "…",
"YELP_API_KEY": "…"
}
}
}
}
Booking is deliberately two-step
resy_book will refuse without confirm: true. The intended flow:
resy_availability→ pick a slot (bookToken)resy_booking_preview→ review time, seating type, and fees, get a realbook_tokenresy_bookwithconfirm: true→ reserved
This makes the one irreversible, money-adjacent action explicit and reviewable.
Safety / honesty notes
- Resy's API is unofficial; endpoints may change. Calls fail loudly rather than guess.
- Nothing here evades bot detection or scrapes rendered HTML.
- Secrets live in
.env/~/.config/reservations-mcp/(gitignored,0600), never in the repo. - Tock is a labelled stub — it does not return fabricated availability.
License
MIT
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。