simsima-connectivity
Public MCP server for a travel-eSIM marketplace: search, recommend and buy data plans for 190+ destinations, returning an attributed checkout link back to simsima.io. Remote Streamable HTTP, no auth.
README
simsima-connectivity — MCP server
Website: simsima.io · Endpoint: POST https://mcp.simsima.io/mcp (Streamable HTTP) · Registry: io.github.Willyfab/simsima-connectivity
Public MCP server exposing Simsima's travel eSIM catalog to AI agents.
Endpoint: POST https://mcp.simsima.io/mcp (Streamable HTTP). Health: GET /health.
Tools
list_destinations— countries/regions covered, with min price + product URL.search_plans— plans for a destination, with optional price/data/validity filters.get_plan— a single plan by sku.recommend_plan— best plan(s) for a trip (length + light/medium/heavy usage).create_checkout_link— a Simsima product URL with agent attribution (source/utm_*).
Read + attributed-link only. No payment, no order data, no auth.
Data source
Reads the public web feed ${SIMSIMA_FEED_BASE}/{locale}/agent/catalog (cached ~15 min).
Product URLs come straight from the feed → always valid (zero 404). No database, no backend calls.
Phase 1 covers country plans in en/fr (regional/global inherit from the feed later).
Env
SIMSIMA_FEED_BASE(defaulthttps://simsima.io)PORT(default 8080)RATE_LIMIT_RPM(default 60)
Local dev
cd mcp
npm install
npm test # jest (25 tests)
npm run dev # ts-node-dev on :8080
curl localhost:8080/health
Connect from Claude (private validation — Phase 1a)
Add a custom connector pointing at https://mcp.simsima.io/mcp, then try:
"Recommande-moi un eSIM pour 7 jours au Japon en usage moyen."
Deploy
Coolify app, Docker build context = mcp/, subdomain mcp.simsima.io. Auto-deploy on push to main.
Set SIMSIMA_FEED_BASE, PORT, RATE_LIMIT_RPM env vars.
Publication to the MCP Registry (Phase 1b)
Metadata lives in server.json (namespace io.github.Willyfab/simsima-connectivity,
remote streamable-http at https://mcp.simsima.io/mcp). Discovery is also hinted from
web/public/llms.txt + llms-full.txt.
To publish (needs the mcp-publisher CLI + GitHub auth as the repo owner):
cd mcp
# install: see github.com/modelcontextprotocol/registry (release binary or `go install`)
mcp-publisher login github # OIDC auth for the io.github.Willyfab/* namespace
mcp-publisher publish # validates server.json and publishes
- Bump
versioninserver.jsonon each meaningful change and re-runpublish. - To use a branded
com.simsima/connectivitynamespace instead, verify domain ownership ofsimsima.iovia a DNS TXT record, then updatenameinserver.json(the remote URLmcp.simsima.ioalready sits on the verified domain).
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。