concerts-paris-mcp
A free public MCP server for searching concerts at the Philharmonie de Paris and Cité de la musique, with filters for genre, price, availability, and more. It provides factual concert details and links to the official booking page.
README
concerts-paris-mcp
A free, public MCP server for searching concerts at the Philharmonie de Paris and the Cité de la musique.
The venue publishes a fine agenda. Its filters are just hard to use for the questions people actually have: baroque, on a weeknight, under 30 euros, still available. This server answers those, then sends you to the official page to book.
Not affiliated. This is an independent project. It is not endorsed by or connected to the Cité de la musique – Philharmonie de Paris. All programme data belongs to them. Tickets are sold only on their site.
Use it
Add the URL to any MCP client that supports remote servers over Streamable HTTP. No account, no API key, no cost.
https://concerts-paris-mcp.samuelbagattin.com/mcp
Tools
| Tool | What it does |
|---|---|
search_concerts |
Free text plus genre, event type, venue, date range, time of day, price ceiling and remaining availability. |
get_concert |
One programme: every date it is performed, the full cast, prices, current availability. Can re-check availability live. |
list_filters |
The controlled vocabulary, so a model can pick valid values instead of guessing. |
catalogue_status |
How complete and how fresh the index is. Worth calling before concluding something does not exist. |
What you get back
Facts, and a link. Date and time in Paris, venue, price range, cast, and one of three availability levels: available, few seats left, sold out. There is no seat count anywhere, because the source does not publish one. Every result states when availability was last checked.
What it deliberately does not do
- No programme notes. The venue's editorial text is indexed so that search works, and is never returned. You get facts and a link to the page that has the prose.
- No images. Not copied, not hotlinked.
- No booking. It cannot reserve, hold or pay for anything.
- No money. No ads, no affiliate links, no data resale, no paid tier.
How it reads the source
The venue has no public API, so the catalogue is built from the same endpoints its own agenda page calls. That earns some obligations, and they are enforced in code rather than promised in a README:
- One request per minute at most, from a single process, with a hard 20-second floor between requests on top of that. A full daily refresh is a few hundred requests, roughly one every four minutes.
- An honest User-Agent, naming this project, its URL and a monitored contact address. The server refuses to start collecting if that address is not configured.
- robots.txt fetched, cached briefly and re-checked before every single request. If it ever disallows these paths, collection stops on its own.
- Exponential backoff,
Retry-Afterhonoured to the second, and a circuit breaker that also trips when responses get slow, on the view that a struggling origin is asking to be left alone. - No evasion of any kind. No IP rotation, no User-Agent spoofing, no working around a block. If the source says stop, it stops.
Those rules are covered by tests in src/source/client.test.ts. If
one of them starts failing, the project is no longer doing what this page says it does.
The legal reasoning behind all of it, including the text and data mining exception it relies on, is written up in docs/LEGAL.md.
Takedown
If you represent the venue and want this switched off, write to the address on the landing page. It will be taken down within 72 hours, no questions asked.
Develop
npm install
cp .dev.vars.example .dev.vars # put a real address in CONTACT_EMAIL
npm run db:migrate:local
npm run dev
npm test # 114 unit tests, no network
npm run check # tsc
Test fixtures are synthetic. They reproduce the shape of the upstream markup, including its awkward corners, without committing any of the venue's own text to this repository.
Deploy your own
npx wrangler d1 create concerts-paris
Put the returned id in wrangler.jsonc, along with your own account_id and route, then:
npm run db:migrate:remote
npx wrangler secret put CONTACT_EMAIL
npm run deploy
CONTACT_EMAIL must be an address you actually read: the collector refuses to send a single
request without one, because an unreachable contact in a User-Agent is worse than none.
/health reports catalogue freshness, guard state and any parser warnings.
How it works
Cloudflare Workers, D1 for the index, one Durable Object for MCP sessions. A cron trigger fires every minute and performs exactly one unit of work: refresh a batch of availability, advance one page of one crawl, or enrich one event page. The cron cadence is the rate limit, and each tick gets a fresh CPU budget. docs/ARCHITECTURE.md has the details.
Licence
Code is MIT, see LICENSE. The data is not: it belongs to the Cité de la musique – Philharmonie de Paris and is neither redistributed nor relicensed here. See NOTICE.
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。