Build a real-time AI classroom by uploading a topic or document
You do not need to create slides or quizzes to explain complex concepts. Enter a topic, and the AI constructs the entire classroom. An AI teacher lectures, AI peers join discussions, and they explain concepts by drawing diagrams or writing formulas on a whiteboard. OpenMAIC is an open-source platform that provides this multi-agent interactive learning environment.
Recent updates: PBL v3 and stability improvements
Version 1.4.1, released on November 5, 2026, fixed issues where unsaved edits were lost during background synchronization or pages added by agents were deleted. It also improved course loading speed on slow networks. The previous version, v1.4.0, introduced Project-Based Learning (PBL) v3, which supports a 'vibe coding' approach where users build projects directly to understand how they work. An AI tutor guides the entire process from requirement definition to implementation, providing milestone feedback and personal learning reports. This feature requires setting up a sandbox provider and a checkpoint repository to activate, and rolling back to previous versions is not possible, so caution is needed before upgrading.
What happens in the classroom
When a user provides a topic or materials, the AI generates a lesson consisting of slides, quizzes, interactive simulations, and project-based learning activities within minutes. The AI teacher and peers converse in real time, drawing diagrams or writing formulas on a whiteboard while explaining via voice. Generated lessons can be exported as PPTX files that retain all scenes, single HTML files playable offline, or ZIP files that can be moved to other deployment environments. It also integrates with various agent workbenches like OpenClaw, Codex, and DeepSeek, allowing classrooms to be created within messenger apps such as Feishu, Slack, and Telegram, or inside IDEs.
Runtime environment and configuration
Running OpenMAIC requires Node.js 22.19 or higher, pnpm 10 or higher, and PostgreSQL 16. Courses are stored on the server, and for local development, you can start the database via Docker using the pnpm db:up command. Model configuration is performed in the openmaic.yml file, where you specify the model for each slot (lecture overview, slide content, text-to-speech, etc.). API keys are stored in .env.local and referenced in the configuration file. The server validates the configuration file at startup to clearly report errors. It supports various providers including OpenAI, Anthropic, Amazon Bedrock, Google Gemini, and DeepSeek, and can also integrate with local AI providers like Lemonade or FunASR.
Deployment considerations
Since v1.2.0, serverless environments (such as Vercel) are not supported; a long-running Node.js process with PostgreSQL is required. Deployment via Docker Compose operates in single-user mode by default, and to expose it to external networks, you must set ACCESS_CODE and change the database password. Security patches have been applied in multiple versions, so using the latest version is important. Specifically, when upgrading to v1.4.0 or higher, all previous version instances must be stopped, and rollback is not possible without a database restore. If you are providing a service for your own product or to others, you must review the owner authentication method and asset quota settings.
