AB
AB
AiBoss

All Categories

project

OpenMAIC - Tsinghua University's open-source multi-agent AI classroom platform

OpenMAIC is an open-source multi-agent AI classroom platform developed by a Tsinghua University team. It can transform any topic or document into an immersive, interactive course with a single click. The platform supports AI teachers giving voice lectures, AI students raising their hands to discuss, and real-time whiteboard drawing, and can generate...

What is OpenMAIC?

OpenMAIC is an open-source multi-agent AI classroom platform developed by a Tsinghua University team. It can transform any topic or document into an immersive, interactive course with a single click. The platform supports AI teachers lecturing via voice, AI students raising their hands to discuss, and real-time whiteboard drawing. It can generate various teaching scenarios, including slides, quizzes, interactive simulations, and project-based learning. The platform is integrated with OpenClaw, allowing users to directly generate courses through chat applications like Lark, reducing preparation costs to only one-thousandth of traditional MOOCs. The system is refined based on two years of real-world classroom data, enabling everyone to access a 24/7 AI classroom.

Main functions of OpenMAIC

  • One-click generationThe platform allows users to input topics or upload documents, and AI automatically generates complete courses that include slides, quizzes, interactive simulations, and project-based learning.
  • Multi-agent classroomAI teachers deliver lessons via voice and operate the whiteboard, while AI students actively ask questions and participate in discussions, creating an immersive learning environment.
  • OpenClaw IntegrationCourses can be generated directly by calling the AI assistant in chat applications such as Lark and Slack, without opening a webpage.
  • Flexible exportThe courseware can be exported as an editable PPTX file or a self-contained interactive HTML file, supporting multiple languages and voice interaction.

OpenMAIC Technical Principles

  • Content generation layerThe platform adopts a two-stage pipeline architecture: In the first stage, the Plan agent analyzes user input and generates a structured course outline. In the second stage, agents in various scenarios work in parallel to transform the outline nodes into specific teaching content, including slide layout, test question design, and interactive web page code generation.
  • Classroom playback engineBased on the Agentic architecture, the AI teacher acts as an autonomous decision-making intelligent agent, driving the teaching process through a state machine. It determines in real time when to give a speech, when to draw on the whiteboard, and when to initiate a discussion, and completes the lesson with the help of TTS speech synthesis and visual effects such as spotlights.
  • Multi-agent orchestration layerUsing LangGraph as the core framework, a Director Graph is constructed to coordinate multiple agent roles and manage the switching of rounds in scenarios such as teacher lecturing, students raising their hands, and roundtable debates, ensuring the orderliness and coherence of multi-agent interactions.
  • Action Execution EngineThe platform encapsulates more than 28 atomic operations, covering voice broadcasting, whiteboard drawing (supporting formulas, charts, and flowcharts), laser pointer pointing, scene switching, etc., providing standardized classroom interaction capability interfaces for upper-level intelligent agents.

OpenMAIC project address

  • Project official websitehttps://open.maic.chat/
  • GitHub repositoryhttps://github.com/THU-MAIC/OpenMAIC

Application scenarios of OpenMAIC

  • Higher EducationUniversity students can quickly generate final exam review courses, such as having AI explain difficult knowledge points like bubble sort and Taylor's formula, and professors can efficiently create professional courseware such as "Introduction to Artificial Intelligence".
  • K-12 Education: Create math olympiad thinking training camps for primary school students, cultivate mathematical thinking through interactive games such as the chicken and rabbit problem and the tree planting problem, or help young students intuitively understand abstract scientific concepts such as general relativity.
  • Lifelong learningThe platform supports users with no prior experience to learn Python programming in 30 minutes, and can instantly analyze the latest DeepSeek papers, interpret stock market trends, and meet other personalized knowledge needs.
  • Corporate TrainingThe platform can quickly create standardized courses such as product knowledge and compliance training, and new employees can learn through role-playing projects to simulate real business scenarios.

Back