Recently, the research team from Tsinghua University officially open-sourced OpenMAIC (Open Multi-Agent Interactive Classroom), a multi-agent interactive classroom platform driven by large language models. This project can instantly transform any topic or document into a complete immersive learning experience, marking the official transition of online education from the "watch video MOOC" era to a new stage of "learning with AI together." The platform currently supports free online experience and has released all code on GitHub, quickly sparking heated discussions in the global AI education community.

image.png

Core Breakthrough: From Single AI Teacher to Complete Virtual Classroom Ecosystem

Traditional AI educational tools have mostly remained at the level of chatbots or automatically generated lecture notes, while OpenMAIC directly builds a "virtual classroom" with multi-role collaboration:

AI Teacher: Responsible for systematic explanations, voice lectures, and real-time drawing and writing formulas on the whiteboard

AI Teaching Assistant: Provides personalized tutoring and detailed supplements

AI Classmate (with various personality prototypes): Simulates real students, actively participates in discussions, asks questions, raises hands, and even engages in roundtable debates

This multi-agent collaboration makes the learning process closer to a real classroom, greatly enhancing interactivity and a sense of companionship.

One-click Generation, Four Teaching Components Automatically Produced

Users only need to input a topic or upload a PDF document, and the system can generate a complete course in about 30 minutes at a very low cost (less than $2), including:

Slide notes

Quizzes and exercises

HTML interactive simulation experiments

Project-based learning (PBL) tasks

The platform integrates OpenClaw capabilities, allowing users to instantly generate and run classes directly in instant messaging tools such as Feishu, Slack, and Telegram, greatly lowering the usage barrier.

Technical Architecture Highlights: Adaptive + Multi-Agent Collaboration

OpenMAIC uses LangGraph to manage multi-agent workflows. Core modules include:

MAIC-Craft: Multimodal content extraction + course planning + automatic generation of agent roles

Adaptive Engine: Achieves token-level personalized teaching based on cognitive student modeling, RAG, Bloom's Taxonomy/ZPD/UDL theories, etc.

Classroom Management Agent: Real-time perception of classroom status and intelligent scheduling of speaking order and discussion rhythm

The project has completed large-scale field trials within Tsinghua University, covering over 500 students and generating more than 100,000 real interaction records, proving its ability to balance scalability and personalization.

A Milestone in AI-native Educational Infrastructure

OpenMAIC is regarded as a major infrastructure innovation in the AI education field. It not only significantly reduces the cost of accessing high-quality educational resources (a smartphone can provide a "Tsinghua-level" full-course private tutor in remote areas), but also offers developers a highly open platform: supports one-click deployment locally, is compatible with mainstream large models (Qwen, GLM, OpenAI, etc.), can be flexibly integrated with TTS/ASR voice services, and welcomes the community to conduct secondary development around course generation, role design, and interaction strategies.

AIbase Review

From the MOOC era to the AI co-learning era, OpenMAIC provides a feasible and scalable technical path. For educators, content creators, and online education platforms, this is an open-source treasure worth in-depth exploration. AIbase will continue to track the community iteration and practical implementation progress of OpenMAIC, bringing more frontline reports.

Experience Address: https://open.maic.chat/

Open Source Repository: https://github.com/THU-MAIC/OpenMAIC