MCP DeepSeek 演示项目
About
MCP DeepSeek 演示项目 is a demo client application built on the Model Context Protocol (MCP) that integrates with the DeepSeek AI model. It demonstrates tool calling and real-time communication between a Next.js frontend and an MCP SSE server.
Details
- Author
- Ulanxx
- GitHub stars
- 21
- Downloads
- 149
- Categories
- Other
Jump to
- Bilingual Chinese/English UI
- Real-time chat with AI assistant
- Supports multiple mock tool calls
- SSE-based real-time communication
- Auto-reconnect mechanism
- Responsive design for mobile & desktop
Clone the repository, install dependencies with pnpm install, configure environment variables (e.g., DEEPSEEK_API_KEY, MCP_SERVER_URL), then run pnpm run dev from the root. Open http://localhost:3000 in your browser; the client auto-connects and shows a chat interface.
MCP DeepSeek Demo Project
<p align="center">
</p>
Certification
Overview
The MCP DeepSeek Demo is a client application built on the Model Context Protocol (MCP), designed to interact with the DeepSeek AI model. This project demonstrates how to integrate the MCP protocol with the DeepSeek API to enable tool calling and real-time communication.
The project follows a monorepo structure with two main components:
- mcp-client: A Next.js-based frontend application providing a user interface and DeepSeek API integration.
- mcp-sse-server: An MCP server using Server-Sent Events (SSE) for real-time tool execution.
Key Features
✅ Modern Chinese/English UI
✅ Real-time chat with AI assistant
✅ Supports multiple tool calls (product lookup, inventory management, etc.)
✅ SSE-based real-time communication
✅ Auto-reconnect mechanism
✅ Responsive design (mobile & desktop)
Getting Started
Prerequisites
- Node.js 16+ & npm/yarn/pnpm
- DeepSeek API Key
Installation
1. Clone the repository
git clone https://github.com/yourusername/mcp-deepseek-demo.git
cd mcp-deepseek-demo
2. Install dependencies
pnpm install
3. Configure environment variables
For mcp-client, create .env and add:
```env
Sign in to leave a review
Use Google, GitHub, or an email account so ratings stay tied to real people.
No reviews posted yet.



