Wegent

🚀 An open-source AI-native operating system to define, organize, and run intelligent agent teams

English | 简体中文

🏗️ Architecture Overview

graph TB
    subgraph Access["Entry Layer"]
        direction TB
        Web["🌐 Web"]
        IM["💬 IM Tools"]
        API["🔌 API"]
    end

    subgraph Features["Feature Layer"]
        direction TB
        Chat["💬 Chat"]
        Code["💻 Coding"]
        Feed["📡 Feed"]
        Knowledge["📚 Knowledge"]
    end

    subgraph Agents["Agent Layer"]
        direction TB
        ChatShell["🗣️ Wegent Chat"]
        ClaudeCode["🧠 Claude Code"]
        Agno["🤝 Agno"]
        Dify["✨ Dify"]
    end

    subgraph Execution["Execution Environment"]
        direction TB
        Docker["🐳 Agent Sandbox"]
        Cloud["☁️ Cloud Device"]
        Local["💻 Local Device"]
    end

    Access --> Features
    Features --> Agents
    Agents --> Execution

✨ Core Features

💬 Chat Agent

• Multi-Model Support: Compatible with Claude, OpenAI, Gemini, DeepSeek, GLM and other mainstream models
• Conversation History: Create new conversations, multi-turn dialogues, save and share chat history
• Group Chat: AI group chat where AI responds based on conversation context with @mentions
• Attachment Parsing: Send txt, pdf, ppt, doc, images and other file formats in single/group chats
• Follow-up Mode: AI asks clarifying questions to help refine your requirements
• Error Correction Mode: Multiple AI models automatically detect and correct response errors
• Long-term Memory: Supports mem0 integration for conversation memory persistence
• Sandbox Execution: Execute commands or modify files via sandbox, E2B protocol compatible
• Extensions: Customize prompts, MCP tools and Skills (includes chart drawing skill)

💻 Code Agent

A cloud-based Claude Code execution engine:

• Multi-Model Configuration: Configure various Claude-compatible models
• Concurrent Execution: Run multiple coding tasks simultaneously in the cloud
• Requirement Clarification: AI analyzes code and asks questions to generate specification documents
• Git Integration: Integrate with GitHub/GitLab/Gitea/Gerrit to clone, modify and create PRs
• MCP/Skill Support: Configure MCP tools and Skills for agents
• Multi-turn Conversations: Continue conversations with follow-up questions

📡 AI Feed

A cloud-based AI task trigger system:

• Full Capability Access: Tasks can use all Chat and Code mode capabilities
• Scheduled/Event Triggers: Set up cron schedules or event-based AI task execution
• Information Feed: Display AI-generated content as an information stream
• Event Filtering: Filter conditions like "only notify me if it will rain tomorrow"

📚 AI Knowledge

A cloud-based AI document repository:

• Document Management: Upload and manage txt/doc/ppt/xls and other document formats
• Web Import: Import web pages and DingTalk multi-dimensional tables
• NotebookLM Mode: Select documents directly in notebooks for Q&A
• Online Editing: Edit text files directly in notebook mode
• Chat Integration: Reference knowledge bases in single/group chats for AI responses

🖥️ AI Device

Run AI tasks on your local machine with full control:

• Local Executor: Install and run the Wegent executor on your own device
• Multi-Device Management: Register and manage multiple local devices
• Default Device: Set a preferred device for quick task execution
• Secure Connection: Connect to Wegent backend via authenticated WebSocket

💬 IM Integration

Integrate AI agents into your favorite IM tools:

• DingTalk Bot: Deploy agents as DingTalk bots for team collaboration
• Telegram Bot: Connect agents to Telegram for personal or group chats

🔧 Customization

All features above are fully customizable:

• Custom Agents: Create custom agents in the web UI, configure prompts, MCP, Skills and multi-agent collaboration
• Agent Creation Wizard: 4-step creation: Describe requirements → AI asks questions → Real-time fine-tuning → One-click create
• Organization Management: Create and join groups, share agents, models, Skills within groups

🔧 Extensibility

• Agent Creation Wizard: 4-step creation: Describe requirements → AI asks questions → Real-time fine-tuning → One-click create
• Collaboration Modes: 4 out-of-the-box multi-Agent collaboration modes (Sequential/Parallel/Router/Loop)
• Skill Support: Dynamically load skill packages to improve Token efficiency
• MCP Tools: Model Context Protocol for calling external tools and services
• Execution Engines: ClaudeCode / Agno sandboxed isolation, Dify API proxy, Chat direct mode
• YAML Config: Kubernetes-style CRD for defining Ghost / Bot / Team / Skill
• API: OpenAI-compatible interface for easy integration with other systems

🚀 Quick Start

One command to start:

curl -fsSL https://raw.githubusercontent.com/wecode-ai/Wegent/main/install.sh | bash

Then open http://localhost:3000 in your browser.

Other Deployment Options

# Standard mode (multi-container with MySQL)
curl -fsSL https://raw.githubusercontent.com/wecode-ai/Wegent/main/install.sh | bash -s -- --standard

# Development mode (from source, with hot reload)
git clone https://github.com/wecode-ai/Wegent.git && cd Wegent && ./start.sh

# Standalone mode (single container)
docker logs -f wegent-standalone      # View logs
docker stop wegent-standalone         # Stop
docker start wegent-standalone        # Start
docker restart wegent-standalone      # Restart

# Standard mode (multi-container)
docker compose logs -f   # View logs
docker compose down      # Stop
docker compose up -d     # Start

# Development mode
./start.sh --status      # Check status
./start.sh --stop        # Stop
./start.sh --restart     # Restart

📖 See Standalone Mode Documentation for details.

📦 Built-in Agents

🤝 Contributing

We welcome contributions! Please see our Contributing Guide for details.

📞 Support

• 🐛 Issues: GitHub Issues
• 💬 Discord: Join our community

👥 Contributors

Thanks to the following developers for their contributions and efforts to make this project better. 💪

Axb	Feifei	MicroLee	FicoHu	YaFei Liu	Yanhe	Johnny0120	Parabala
Icycrystal4	Yi Xiang	Joyway78	Moqimoqidea	Xuemin	Jake Zhang	Fengkuizhi	Kerwin Bryant
Just Quan	Junbaor	Fingki	Birch	李枨煊	Jolestar	Code-wangdi	Sense_wang
Li Daiyan	Erdawang	Jasper Zhong	Ray	RichardoMu	RockysGit	_	Andrewzq777
Graindt	Qingcheng	Salt-hai	Wxcfox

Made with ❤️ by WeCode-AI Team