OpenMAIC — Multi-Agent Interactive Classroom for Immersive Learning

Introduction

OpenMAIC (Open Multi-Agent Interactive Classroom) is a research-driven platform that simulates classroom environments using multiple AI agents. Each agent assumes a distinct role — lecturer, teaching assistant, student, or peer reviewer — creating dynamic educational interactions that go beyond single-prompt Q&A.

What OpenMAIC Does

• Simulates multi-agent classroom sessions with configurable roles
• Supports Socratic dialogue, debate, and peer-review workflows
• Provides a web UI for observing and participating in agent discussions
• Allows custom curriculum and lesson plan injection
• Exports session transcripts and learning analytics

Architecture Overview

OpenMAIC uses a TypeScript orchestrator that manages agent lifecycles and message routing. Each agent runs as an independent context with its own system prompt and memory. The orchestrator enforces turn-taking rules and topic progression according to a lesson plan graph. The frontend renders a classroom view where users can watch agent interactions in real time or inject their own questions.

Self-Hosting & Configuration

• Requires Node.js 18+ and an LLM API key (OpenAI, Anthropic, or local Ollama)
• Classroom scenarios are defined in YAML lesson plan files
• Agent roles and personalities are configured via JSON persona templates
• Supports PostgreSQL for persistent session storage or SQLite for quick setup
• Docker Compose file included for one-command deployment

Key Features

• Multi-agent role-play with realistic classroom dynamics
• Configurable teaching strategies (Socratic, lecture, group discussion)
• Real-time web UI with chat-style classroom view
• Session recording with exportable transcripts
• Extensible agent persona system for custom domains

Comparison with Similar Tools

• AutoGen — general multi-agent framework; OpenMAIC is purpose-built for education
• CrewAI — task-oriented agent teams; OpenMAIC focuses on dialogue-driven learning
• ChatGPT — single-agent conversation; OpenMAIC orchestrates multiple agents simultaneously
• Coursera/edX — pre-recorded content; OpenMAIC generates interactive sessions on demand

FAQ

Q: Can I use local models instead of cloud APIs? A: Yes, OpenMAIC supports Ollama and any OpenAI-compatible local endpoint for fully offline operation.

Q: What subjects work best? A: It performs well in STEM, programming, and humanities. Custom curricula can target any domain with appropriate persona configuration.

Q: How many agents can run in a single session? A: The default setup handles 4-6 agents comfortably. Scaling beyond that depends on the LLM provider rate limits and context window size.

Q: Is this suitable for production educational platforms? A: OpenMAIC is primarily a research tool. Production deployment would require additional auth, rate limiting, and content moderation layers.

Sources

• https://github.com/THU-MAIC/OpenMAIC