Magic: An Open-Source AI Productivity Platform with Agent Automation

Magic is an open-source, all-in-one AI productivity platform designed to streamline complex business processes. The ecosystem consists of four primary modules: a general-purpose AI agent, a robust workflow engine, a centralized knowledge base, and a suite of online collaboration tools.

1. Super Magic

Super Magic is a sophisticated AI agent engineered to manage intricate tasks. Built on a multi-agent architecture and equipped with an extensive toolset, it interprets natural language requests, formulates action plans, executes steps, and autonomously corrects its own errors. By processing natural language inputs, it manages business workflows from start to finish. As an open-source flagship product, Super Magic allows organizations to customize the agent for specific operational needs, leading to faster decision-making and improved outcomes.

Current applications of Super Magic include:

• Analyzing insights from Warren Buffett’s 2025 shareholder meeting.
• Researching stocks related to the Beijing humanoid robot half-marathon.
• Summarizing core concepts from Thinking, Fast and Slow.
• Providing IPO analysis and investment notes on "Aunt Jenny."
• Forecasting sales demand at the SKU level.

2. Magic IM

Magic IM is a conversational system tailored for internal knowledge management and customer support. It manages multi-turn dialogues, maintains context throughout a conversation, and queries the knowledge base to provide accurate information. Organizations use it to rapidly deploy intelligent Q&A bots and knowledge assistants.

Core Features:

• Knowledge Base Management: Users can upload documents in various formats. The system indexes this data for semantic search, ensuring that AI responses are grounded in the company’s specific data.
• Conversation Management: The platform keeps discussion threads organized, handling both AI agent interactions and human-to-human internal messaging.
• Group Chat: Teams can engage in real-time discussions with an AI assistant present in the chat to provide immediate answers and facilitate knowledge sharing.
• Multi-Organization Support: Each organization operates within its own secure environment, ensuring data isolation and strict permission management.
• Data Security: Layered access controls and rigorous isolation protocols protect sensitive material across organizational boundaries.

3. Magic Flow

Magic Flow provides a visual workspace for building AI agent workflows through an intuitive drag-and-drop interface that requires no coding.

Core Features:

• Visual Builder: A user-friendly canvas where nodes can be linked to design complex logic and automation.
• Component Library: Includes pre-built nodes for text processing, image generation, code execution, and more.
• Broad Model Support: Compatible with any large language model (LLM) that adheres to the OpenAI API protocol, allowing users to select the best "brain" for their specific task.
• System Integration: Connects directly with Magic IM and external platforms like WeCom, DingTalk, and Lark for frictionless cross-platform collaboration.
• Custom Extensions: Developers can write proprietary tool nodes for niche requirements.
• Live Debugging and Monitoring: Offers full observability into workflows, allowing users to identify bottlenecks and resolve errors before they impact production.

4. Teamshare OS (Coming Soon)

Teamshare OS is a modern enterprise collaboration hub where AI is woven into daily office tasks to organize knowledge and streamline workflows.

Core Features:

• Smart Document Management: Provides collaborative editing for various file formats, complete with version control and AI-assisted drafting and polishing.
• Magic Table: A flexible data tool featuring custom fields, multiple views, and automated routines, with AI-driven data processing.
• Project Collaboration: Includes Kanban boards and task tracking with adjustable workflows. AI monitors progress to flag potential resource issues before they escalate.
• Knowledge Base: Consolidates internal documents into a structured, searchable asset that scales with the company.
• Full Integration: Connects the entire Magic product family while plugging into standard office software to create a unified workspace.

Open-Source Infrastructure

• Agentlang: A language-first framework that allows users to build AI agents using natural language.
• Magic Lens: A high-performance HTML-to-Markdown converter that uses flexible rules to transform complex web pages into clean Markdown.
• Magic Use: A browser-based toolset specifically designed for AI agent interaction.
• Magic Space: A management system for hosting static content used by AI agents.
• Sandbox OS: A secure, high-performance environment where AI agents can execute tasks safely.

How to Use Magic

Cloud Service

Super Magic, Magic IM, and Magic Flow are all available via cloud access with no setup required. The cloud version includes every feature found in the open-source edition. Currently, access is granted via invitation; users can submit a request online for approval to begin a trial.

Enterprise Customization

For teams and companies requiring advanced administrative tools and tailored services, the official Magic team provides specialized support via email inquiry.

Self-Hosted Community Edition

1. System Requirements

• Docker 24.0 or newer
• Docker Compose 2.0 or newer

2. Deployment Clone the repository:

git clone https://github.com/dtyq/magic.git
cd magic

Start the services:

• Foreground: ./bin/magic.sh start
• Background (Daemon): ./bin/magic.sh daemon

3. Other Useful Commands

• Check status: ./bin/magic.sh status
• View logs: ./bin/magic.sh logs

4. Environment Variables

• Magic variables: Copy .env.example to .env and configure at least one LLM variable.
• Super Magic variables: Copy config/.env_super_magic.example to .env_super_magic and set up an OpenAI-compatible model.

5. Access Points

• API Service: http://localhost:9501
• Web App: http://localhost:8080 (Default credentials: 13812345678 / letsmagic.ai and 13912345678 / letsmagic.ai)
• RabbitMQ Dashboard: http://localhost:15672 (Username: admin, Password: magic123456)

Magic Website: www.letsmagic.ai
Magic Documentation: docs.letsmagic.cn/en