THU-MAIC/OpenMAIC

OpenMAIC – Open Multi‑Agent Interactive Classroom

What it is – OpenMAIC is an open‑source web platform that lets you turn any topic, document, or set of learning materials into a fully‑featured, AI‑driven classroom. It orchestrates several AI agents (teacher, classmates, whiteboard assistant, TTS voice, etc.) to generate slides, quizzes, interactive HTML simulations, and project‑based learning activities, and then delivers them in real time through a browser UI.

Key capabilities

• One‑click lesson generation – describe a subject or upload source files and the system builds a complete lesson (outline, slide content, images, audio narration, optional video) automatically.
• Multi‑agent interaction – AI “teachers” and “peers” can speak, draw on a shared whiteboard, answer questions, and run simulations, giving a conversational learning experience.
• Rich media output – export editable PowerPoint (.pptx) decks or self‑contained interactive HTML pages; media can be stored in PostgreSQL or S3.
• Agent‑workbench integration – the built‑in OpenMAIC Skill works with OpenClaw, Codex, DeepSeek, WorkBuddy and many other agent frameworks, so classrooms can be generated from Feishu, Slack, Telegram, or directly inside an IDE.
• Pluggable model providers – supports OpenAI, Anthropic, Azure OpenAI, Amazon Bedrock, Gemini, Qwen, MiniMax, Grok, TokenDance, Xiaomi MiMo, Zhipu GLM, Ollama, Lemonade (local) and any OpenAI‑compatible API. Model usage is configured via a simple openmaic.yml file that maps slots (outline, slide text, TTS, web search, etc.) to specific providers.
• Server‑backed persistence – courses, chat history, generated media and user sessions are stored in PostgreSQL, allowing the generation process to survive closed tabs or server restarts.
• Deployable via Docker or locally – development uses pnpm and a Docker‑based PostgreSQL; production can run as a long‑living Node.js process (Docker Compose) or, for older 1.1.x releases, on Vercel serverless.

Typical use cases

Scenario	How OpenMAIC helps
Educators building quick, interactive lessons for a class or workshop.	Generate slides, quizzes and live simulations in minutes; export to PowerPoint for offline use.
Corporate trainers needing on‑the‑fly training modules.	Upload product docs, let the AI create a structured course with voice‑over and interactive demos.
Students wanting a personal tutor for a textbook chapter.	Describe the chapter, get a multi‑agent classroom that can answer follow‑up questions and draw explanations.
Developers / IDE users who want to embed learning into their workflow.	Use the OpenMAIC Skill in OpenClaw or other workbenches to spin up a lesson directly from a messaging channel or code editor.
Researchers experimenting with multi‑agent orchestration.	The open‑source stack (Next.js, LangGraph, Tailwind) is fully accessible for extending agents or adding new modalities.

Quick start (local development)

# 1. Clone the repo
git clone https://github.com/THU-MAIC/OpenMAIC.git && cd OpenMAIC

# 2. Install dependencies (pnpm ≥10)
pnpm install

# 3. Set up configuration
cp .env.example .env.local            # put your API keys here
cp openmaic.example.yml openmaic.yml  # edit model slots if desired
# At minimum you need OPENAI_API_KEY (or another provider key)

# 4. Start a PostgreSQL instance (Docker helper)
pnpm db:up
# The helper creates a DB at postgres://openmaic:openmaic-dev@127.0.0.1:5432/openmaic

# 5. Run the dev server
pnpm dev

# Open http://localhost:3000 in a browser and start generating lessons.

Production / Docker deployment

# Prepare env file with keys
cp .env.example .env.local && nano .env.local

# Build and run both the app and PostgreSQL
docker compose up --build -d

The compose stack creates two containers (app + Postgres) and persists data in named volumes (openmaic-postgres, openmaic-data). For multi‑user deployments set an ACCESS_CODE and a strong PERSISTENCE_POSTGRES_PASSWORD before publishing.

Where to learn more

• User guides – English and Chinese guides are linked at the top of the README.
• Documentation – detailed docs live under packages/docs/content/docs/ (configuration, supported models, deployment, hosting, etc.).
• Live demo – https://open.maic.chat/ shows the UI in action.
• Community – Discord and Feishu links are provided for support and discussion.

---

OpenMAIC is a genuine, actively maintained AI project focused on multi‑agent classroom generation, model orchestration, and interactive learning experiences. It fits squarely within the frontier AI/ML space (LLM orchestration, RAG, multimodal generation, and AI‑driven education).