Super McGee: An Enterprise-Grade Open-Source AI Agent Platform 200

What is Super Magic

Super Magic is an enterprise-grade open-source AI Agent platform developed by Guangdong Lighthouse Engine Technology Co., Ltd. It is positioned as an "all-in-one AI productivity engine." It is not a chat bot, but rather a general-purpose AI Agent system designed for complex tasks. It supports multi-Agent collaboration, visual workflow orchestration, and can integrate with enterprise IM systems. Super Magic is based on an Apache 2.0-derived license, supports private deployment, and aims to let enterprises "gain the execution power of a hundred people at the cost of one." It is suitable for team productivity improvement, sales automation, and complex research and analysis scenarios.

Super Magic official website screenshot

Features

• Multi-Agent Intelligent Collaboration: It can autonomously understand tasks, plan, execute, and self-correct. Multiple Agents can work together as a team—for example, a research Agent, a content Agent, and a design Agent working together to complete a full report.

• Visual Workflow Orchestration (Magic Flow): A drag-and-drop, zero-code interface that includes atomic nodes for text processing, image generation, code execution, and more. It can connect with enterprise tools such as WeCom, DingTalk, and Feishu.

• Results as Deliverables: Through its underlying rendering framework, it directly generates PPTs, data dashboards, professional reports, Excel files, meeting minutes, and other ready-to-use deliverables, rather than merely providing text suggestions.

• Enterprise-Grade Security and Compliance: A self-developed sandbox container provides isolation (multi-tenant VPC isolation), operational audit logs, and high-risk operations require manual approval, ensuring data stays within a trusted boundary.

• Granular Cost Compass: Set daily budget caps by department, user, or Agent, making AI spending 100% traceable and controllable.

• Open Ecosystem Compatibility: Compatible with the Anthropic Skills and OpenClaw Skills ecosystems, and supports the MCP protocol for integrating third-party services.

• Expert Agent Marketplace: Comes with built-in digital employees for research and analysis, data processing, content generation, meeting insights, and design processing—ready to use out of the box.

Use Cases

• One-Person Companies or Lean Teams: Replaces roles such as marketing, operations, legal, and data analysis, producing dashboards, draft contracts, and advertising creatives 24/7.

• End-to-End Sales Efficiency Boost: Automatically pushes customer follow-up to-dos, analyzes historical meeting recordings and generates interactive dashboards, automatically applies templates to produce quotations, and synchronizes follow-up status updates in the CRM.

• Enterprise Collaboration: Can serve as an alternative to Feishu/DingTalk, supporting AI-powered knowledge retrieval in group chats, parallel progress on the same project by multiple people, and automatically reporting results to IM groups.

• Complex Research and Analysis: Batch-generates industry research reports, individual stock analysis, and sales forecasts, and supports simultaneously processing hundreds of documents.

• Security and Surveillance: Integrates with RTSP cameras to perform face detection and tracking, and uploads images to the cloud.

Pros and Cons of Super Magic

• Pros: Open-source and free, with support for secondary development; enterprise-grade security architecture (sandbox isolation, private deployment, audit logs); multi-Agent collaboration and project-based management; directly produces usable documents rather than plain text; granular cost control, suitable for enterprise budget management.

• Cons: There are functional differences between the open-source community version and the official latest product, with some advanced tools not being synced to the community version; the maturity of the plugin ecosystem relies on community maintenance; deep customization and on-premise deployment have a certain technical threshold for the technical team; outputs in complex business scenarios (such as finance and legal) still require manual review.

How to Use Super Magic

• Quick Cloud Experience: Go to the official website at https://www.letsmagic.cn to register an account, and you can use the cloud versions of Super Magic, Magic Flow, and Magic IM.

• Private Deployment: Make sure Docker 24.0+ and Docker Compose 2.0+ are installed in your environment, then run the following command for a one-click deployment: curl -fsSL https://getmagicrew.sh | bash

• Once deployment is complete, access it via http://localhost:38080.

• Create an Agent: Use natural language instructions or drag-and-drop nodes in the visual interface to define the Agent's capabilities, configure the large model API (supporting OpenAI, Claude, Kimi, Qwen, etc.), and then you can start orchestrating tasks.

FAQ

• What are the deployment resource requirements? For basic services, 2 vCPU / 4 GB RAM is recommended; for production environments, 4 vCPU / 8–16 GB RAM is recommended. Additional GPU or CPU resources are needed if connecting to a local large model.

• How is data security ensured? Full private deployment is supported. Agents run in isolated sandbox containers that are not in the same VPC as the main system, multi-tenant data is completely isolated, and there are full operational audits.

• How is Super Magic different from Coze and Dify? Super Magic focuses more on enterprise-grade collaboration, security and compliance, result delivery, and cost control. It comes with built-in IM and office collaboration capabilities, rather than focusing solely on single-point AI application development.

• Does it support domestic large models? It supports all models that follow the OpenAI API protocol, including domestic models such as Kimi and Qwen, which can be switched with one click.

Review

The core differentiation of Super Magic lies in its positioning as an "enterprise-grade premium digital employee." Unlike personal assistants such as ChatGPT and Claude, it emphasizes project-based collaboration, knowledge accumulation that does not leave with the person, and "delivering finished results rather than semi-finished products." 36Kr commented that it targets the deep-water enterprise areas that traditional AI Agents cannot enter—making AI ready-to-use under the premise of absolute security and controllability. Real-world feedback shows significant efficiency improvements in scenarios such as batch report generation and front-end page design, and the design aesthetics have also received positive user reviews.