mcp-starter-kit
A starter kit for creating a read-only MCP server that lets AI assistants access business information, calculate prices, answer FAQs, and check availability.
README
🐇 The Plain-English MCP Starter Kit
Everyone's talking about "MCP." Here's one you can actually use — in 5 minutes, with your own business.
You've probably heard that AI can now "connect to your tools." That connection has a name: MCP (Model Context Protocol). It's how an AI assistant stops guessing and starts answering with your real business data — your hours, your prices, your FAQs.
This kit is a tiny, working MCP server. You edit one file with your info, run one command, and connect it to your AI. Then you can ask your AI things like:
"How much for 3 lattes and a pastry platter?" "What are your Saturday hours?" "What times are open this week?"
…and it answers correctly, every time — because it's reading your data, not making it up.
💬 Want the link to this kit? Comment "MCP" on the post and we'll send it over.
What's an MCP, in one sentence?
An MCP server is a safe translator between your AI and your stuff. The AI can only do the specific actions you allow — nothing else on your computer.
See the picture: docs/diagram.md
This kit allows exactly four safe, read-only actions:
| Action | What it does |
|---|---|
get_business_info |
Shares your hours, contact, and address |
price_quote |
Calculates a price from your service list |
lookup_faq |
Answers common customer questions |
check_availability |
Lists your open booking slots |
Set it up in 5 minutes
You need: Node.js (the free "LTS" download) and a free AI app that supports MCP, like Claude Desktop. You do not need to know how to code.
1. Download this kit
Click the green Code button above → Download ZIP → unzip it.
2. Add your business info
Open business.json in any text editor and replace the example
(Maple Street Coffee) with your own name, hours, services, prices, and FAQs.
Save it.
3. Install + test it
Open Terminal, go into the folder, and run:
npm install
npm run check
You should see ✅ Your kit looks good!
4. Try it in your browser first (the fun part) 🎮
Before touching any AI app, run:
npm run play
Then open http://localhost:4321. You'll get a simple page where you can:
- Edit your business info in a form (no JSON, no code) and Save
- Try the tools live — ask "What are your hours?", get a price quote, look up an FAQ — and see the exact answers your AI will give
This is the same logic the AI uses, so if it works here, it works there.
Press Ctrl+C in Terminal to stop the playground.
5. Connect it to your AI
In Claude Desktop: Settings → Developer → Edit Config, and add this (change the path to where you unzipped the folder):
{
"mcpServers": {
"my-business": {
"command": "node",
"args": ["/full/path/to/mcp-starter-kit/server.js"]
}
}
}
Restart Claude Desktop. You'll see your business tools appear. Now ask it a question about your business — and watch it use the real answer. 🎉
The 5 questions to ask before you connect AI to anything
Before you wire AI into your real tools (email, CRM, calendar, payments), ask:
- What can it actually do? — Read-only, or can it change/delete things? (This kit is read-only on purpose.)
- What data does it touch? — And is any of it sensitive?
- Where does that data go? — Stays on your machine, or sent to a company?
- Who can use it? — Just you, or your whole team / your customers?
- What happens if it's wrong? — What's the worst-case, and can you undo it?
If you can't answer these for a tool someone's trying to sell you — slow down.
Make it yours
Everything lives in business.json — no coding needed to update your info.
Want it to do more (real bookings, live inventory, connect to your CRM or
email)? That's exactly the kind of thing we build.
🐇 Built by Rabbithole
We help small and medium businesses put AI to work — safely, and in plain English. This is Day 1 of our 30-day series of free, working AI kits.
Want this set up for your business, with your real tools connected? 👉 Reach out to Rabbithole (add your link)
Free to use under the MIT License. Built to be opened, edited, and run by real business owners.
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。