HyperChat

中文 | English

Introduction

HyperChat is an open Chat client that can use APIs from various LLMs, fully supports MCP, and provides the best Chat experience. It also implements productivity MAX tools based on native Agents.

• Supports OpenAI-style LLMs, OpenAI, Claude(OpenRouter), Qwen, Deepseek, GLM, Ollama.
• Built-in MCP plugin market, user-friendly MCP installation configuration, one-click installation, welcome to submit HyperChatMCP.
• Also supports manual installation of third-party MCPs, just fill in command, args, env.

DEMO

• HyperChat on Docker

Features:

• [x] 🪟Windows + 🍏MacOS + Linux
• [x] Command line run, npx -y @dadigua/hyper-chat, default port 16100, password 123456, Web access http://localhost:16100/123456/
• [x] Docker
  • Command line version docker pull dadigua/hyper-chat-mini:1.2.8
  • Ubuntu desktop + Chrome + BrowserUse version (coming soon)
• [x] MCP extension market + third-party MCP support
• [x] Supports free creation, design of Agent, can preset prompts, choose MCP functions
• [x] Supports dark mode🌙
• [x] Supports Resources, Prompts, Tools
• [x] Supports English and Chinese
• [x] Supports Artifacts, SVG, HTML rendering, JS error capture, supports opening Chrome console
• [x] Agent display optimization, supports search, drag-and-drop sorting
• [x] Supports KaTeX, displays mathematical formulas, code rendering increases highlighting and quick copy
• [x] WebDAV synchronization
• [x] Added RAG, based on MCP knowledge base
• [x] Introduced the concept of ChatSpace, supports simultaneous chatting with multiple dialogues
• [x] Supports Agent Call Agent through HyperAgent’s MCP
• [x] Added scheduled tasks, specifying Agents to complete tasks at set times, and viewing task completion status.

TODO:

• [ ] Future updates will include a docker version with a built-in Linux desktop, built-in Chrome configuration, unified environment for easier handling, and web interface control. It can be used on any device, including mobile phones. 🤣
• [ ] Permission pop-up, whether to allow
• [ ] Implement LLM writing MCP for itself

LLM

| LLM | Usability | Remarks | | -------- | -------- | -------------------------- | | claude | ⭐⭐⭐⭐⭐ | No explanation | | openai | ⭐⭐⭐⭐⭐ | Also perfectly supports multi-step function calls (gpt-4o-mini can also) | | gemini flash 2.0 | ⭐⭐⭐⭐🌙 | Very useful | | qwen | ⭐⭐⭐⭐🌙 | Very useful | | doubao | ⭐⭐⭐🌙🌙 | Feels okay to use | | deepseek | ⭐⭐⭐🌙🌙 | Multi-step function calls may have issues |

Usage

• 1. Configure APIKEY, ensure your LLM service is compatible with OpenAI style.
• 2. Make sure your system has uv + nodejs installed.

uvx & uv

Install using the command line, or check the official GitHub tutorial uv

# MacOS
brew install uv
# Windows
winget install --id=astral-sh.uv  -e

npx & nodejs

Install using the command line, or go to the official site to download, official site nodejs

# MacOS
brew install node
# Windows
winget install OpenJS.NodeJS.LTS

Development

cd electron && npm install
cd web && pnpm install
npm install
npm run dev

Notes

• MacOS may encounter damage or permission issues, sudo xattr -d com.apple.quarantine /Applications/HyperChat.app
• MacOS nvm users manually input PATH echo $PATH, Windows version nvm seems to work directly

Telegram

HyperChat User Communication

Call shell mcp

Call terminal mcp, ssh + can execute commands

One-click to write a webpage and publish to (cloudflare)

Call Google Search, ask it what TGA Game of the Year is

Organize Zhihu hot searches

Help you open web pages, analyze results, and write to files

Open Baidu and take screenshots

Scheduled task list

MCP market (experimental)

Install MCP interface from the market (experimental)

Install MCP from third-party (supports any MCP)

Install MCP interface from third-party

MCP list (can be dynamically selected)

Render HTML, support Artifacts, SVG, HTML rendering,

Interface 1

Interface 2

Interface 3, test model capabilities

Knowledge base

Disclaimer

• This project is for learning and exchange purposes only. If you use this project for any actions, such as crawling behaviors, it has nothing to do with the developers of this project.