MCP Elicitation and Apps Demo
MCP server demonstrating elicitation for booking appointments and an interactive MCP App for viewing appointments with SQLite persistence.
README
MCP Elicitation and Apps Demo
A minimal MCP server demo with Elicitation and MCP Apps support for webinar presentations.
Features
- Elicitation Tool:
book_appointmentwith SQLite persistence - MCP App:
view_appointmentswith interactive HTML UI - Tailwind CSS: Modern, responsive interface
- Real-time Data: Live appointment statistics
- SQLite Database: Data persistence and sample data
- HTTP Transport: Endpoint at
http://localhost:3000/mcp
MCP Apps Feature
This demo now includes MCP Apps support with:
- Interactive HTML interface rendered in MCP hosts
- Tailwind CSS for modern styling
- Real-time appointment data visualization
- Statistics dashboard
- Sample data pre-loaded
Database Schema
CREATE TABLE appointments (
id INTEGER PRIMARY KEY AUTOINCREMENT,
person TEXT NOT NULL,
duration_minutes INTEGER NOT NULL,
created_at TEXT NOT NULL
);
Installation
# Install dependencies
npm install
# Build TypeScript
npm run build
Running the Demo
Elicitation Demo (Original)
# Start the original elicitation server
npm start
# Or for development with auto-reload:
npm run dev
MCP Apps Demo (New)
# Start the MCP Apps server
npm run start:apps
# Or for development with auto-reload:
npm run dev:apps
Simple Test
To see the expected flow:
npm run test
Webinar Flow
- Start server:
npm start - Connect an MCP client to
http://localhost:3000/mcp - Ask: "Book an appointment with John"
- Elicitation asks for duration: 30 / 60 / 90 minutes
- Choose 60
- Open
appointments.dbin a SQLite GUI - Show the inserted row with John, 60 minutes, and timestamp
Tool Specifications
Book Appointment (Elicitation Demo)
Name: book_appointment
Initial input:
{
"person": "string"
}
Elicitation (when duration is missing):
- Message: "How long should the appointment be?"
- Schema: Enum of 30, 60, or 90 minutes
Result: Inserts into SQLite and returns confirmation message.
View Appointments (MCP App)
Name: view_appointments
Input: None required
Result: Opens an interactive HTML interface with:
- List of all appointments
- Statistics dashboard (total appointments, total duration, average duration)
- Real-time data refresh
- Modern Tailwind CSS styling
MCP App Features:
- Interactive UI rendered directly in MCP hosts
- HTML content served via MCP resources API
- Real-time data from SQLite database
Code Structure
Original Demo
src/server-simple.ts- Simple MCP server with Elicitation logicsrc/db.ts- SQLite database operationsappointments.db- SQLite database file (created on first run)
MCP Apps Demo
src/server-apps.ts- Extended server with MCP Apps supportsrc/appointments-ui.html- Interactive HTML interface with Tailwind CSSappointments.db- Shared SQLite database with sample data
Key Implementation Points
- Tool starts - Receives
personparameter - Elicitation - Detects missing
duration_minutes, returnsinputRequired - Continue - After user provides duration, completes operation and saves to DB
Testing
You can test both servers using any MCP-compatible client (Claude Desktop, VS Code with MCP extension, etc.) by connecting to http://localhost:3000/mcp.
Testing Elicitation Demo
- Connect to
http://localhost:3000/mcp - Call
book_appointmentwith{"person": "John"} - Server will elicit for duration (30/60/90 minutes)
- Provide duration to complete the appointment
Testing MCP Apps Demo
- Start the MCP Apps server:
npm run start:apps - Connect to
http://localhost:3000/mcp - Call
view_appointments(no parameters needed) - MCP host will render the interactive HTML interface
- Explore appointments with statistics and real-time data
Requirements
- Node.js 18+
- SQLite GUI (for viewing the database during demo)
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。