OpenMAIC

OpenMAIC is the open-source release of the MAIC multi-agent classroom described in the MAIC study: describe a topic or attach your own materials and it generates a full lesson — slides, quizzes, interactive HTML simulations and project-based activities — then delivers it through AI teachers and AI classmates who speak, draw on a whiteboard and take part in discussion. It is the code behind a classroom that can be run rather than only read about.

What you can do with it

One-click generation produces a lesson in minutes from a prompt or uploaded document, audio or video. The multi-agent layer adds classroom discussion the learner can join or be called into, roundtable debates between personas with whiteboard illustrations, and free-form Q&A where the teacher answers with slides or diagrams. Sessions support slide decks, quizzes, interactive simulations and PBL, and export as editable .pptx or interactive .html. Version 1.0.0 (August 2026) added an agent workbench: a chat-first workspace that plans and revises whole courses, durable server-backed sessions you can cancel, resume or steer, and 24 built-in skills covering slides, quizzes, interactives, images, video and voices. A SKILL.md package lets an agent harness build classrooms from a messaging app.

Who it is for

Instructors and course teams who want generated materials they can still edit, instructional designers prototyping multi-agent activities, and developers who need a self-hostable classroom rather than a hosted product. Schools can deploy it on Vercel or with Docker, and the project ships a hosted demo at open.maic.chat.

Notes

Licensed MIT, with an English and Chinese user guide and an active community on Discord and Feishu. It is model-neutral: you supply at least one LLM provider key, and optional local components (Lemonade for local models, FunASR for speech recognition) let you run more of the stack offline — so "free" describes the software, not the inference bill. The paper behind it appeared in the Journal of Computer Science and Technology (2026, DOI 10.1007/s11390-025-6000-0), and the repository had more than 38,000 stars by September 2026.

Connected Concepts

Agentic AI, Generative AI, Large Language Models (LLMs), Open Source, Project-Based Learning, Online Teaching and Learning, Personalized Learning, Teaching