content / agent

OpenMAIC

An open-source AI classroom platform for topic- or document-based lessons, with slides, quizzes, simulations, AI teachers and AI classmates.

Repository owner THU-MAIC

Not rated yet
0 comments

Like 0 Save

Published 11 Oct 2026Listing updated 12 Oct 2026

At a glance

Purpose: An open-source AI classroom platform for topic- or document-based lessons, with slides, quizzes, simulations, AI teachers and AI classmates.
Runs in: Chat app with skills, Local agent with shell, Web service
Operating systems: Not stated
Needs: Runtime: Node.js 22.19+, pnpm 10+, PostgreSQL 16; Needs internet: Yes; Needs accounts or API keys beyond your agent: Not stated
Files and data: Reads files: Yes; Writes files: Yes; Sends data to other services: Yes
Tool type: Agent
Delivery: From GitHub
Version: Latest on the main branch
Access: Code on GitHub
Compatibility: According to the README, OpenMAIC runs on Node.js with PostgreSQL, and it ships a Docker Compose setup. It lists providers including OpenAI, Azure OpenAI, Anthropic, Amazon Bedrock, Google Gemini, DeepSeek, Qwen, Kimi, MiniMax, Grok (xAI), OpenRouter and Ollama, any OpenAI-compatible API, and local Lemonade and FunASR. The OpenMAIC Skill is described as working with OpenClaw, Codex, DeepSeek and WorkBuddy, and classrooms can be generated from Feishu, Slack and Telegram. The README says Vercel deployment is supported only up to 1.1.x.
External costs: The README names many commercial model providers, including OpenAI, Anthropic, Amazon Bedrock and Google Gemini, and token plan providers such as MiniMax and TokenDance, which it says can cover chat, image, video, speech and web search with one key. It names a MinerU official API key for document parsing. It says Lemonade and FunASR need no API key. The repository is MIT-licensed. The model provider you connect may charge for usage.
Permissions: The README describes installing the OpenMAIC Skill with clawhub install openmaic or importing skills/openmaic/ into another workbench. The README says the skill submits generation jobs, polls their progress and sends the resulting link. The README's self-hosted setup uses API keys that you add to .env.local. The repository file list includes scripts/, whose behavior is not described in the README. The README says the workbench, off by default, can create courses and folders; rename and move courses; edit, insert, delete and reorder pages; upload and extract materials; fetch web content; and generate media through configured providers. The README describes microphone input and voice clips sent to a configured VoxCPM backend. The README says generation runs in a long-running server and continues after the browser leaves the page.
Usage terms: MIT License, as stated in the repository. See its LICENSE file for the exact terms. The README says packages/mathml2omml retains LGPL-3.0-or-later terms and packages/pptxgenjs retains its own MIT terms. The README says each bundled package's terms apply to its files when redistributing the repository as a whole.
Interface: web

Code on GitHub

THU-MAIC/OpenMAIC

MIT

TypeScript40,538 stars6,277 forksLast pushed 12 Oct 2026 UTC

Open Multi-Agent Interactive Classroom — Get an immersive, multi-agent learning experience in just one click

Public GitHub data fetched 13 Oct 2026 UTC · 0 days old

Skill folders in this repository (25)

Folders that contain a SKILL.md file, from GitHub's file list on 13 Oct 2026 UTC.

View on GitHub ↗

What it does

The README describes OpenMAIC, Open Multi-Agent Interactive Classroom, as an open-source AI classroom platform. The README says its multi-agent orchestration generates slides, quizzes, interactive simulations and project-based learning activities. The README says AI teachers and classmates deliver lessons, speak, draw on a whiteboard and discuss with the learner. The README lists editable .pptx slides and interactive .html pages as exports.

The README describes an OpenMAIC Skill for OpenClaw and agent workbenches such as Codex, DeepSeek and WorkBuddy. The README says this integration supports classroom generation from messaging apps such as Feishu, Slack and Telegram or from an IDE.

The README says the server stores courses in PostgreSQL and uses openmaic.yml to declare providers and assign models to AI capabilities. The README labels v1.2.0-rc.1 a pre-release and says generation runs on the server and continues after a browser tab closes.

Who it is for

People who want AI-led lessons built from a topic or their own materials, and developers who host the platform for themselves or others.

Example tasks

Describe a topic or attach your materials to generate a lesson.
Discuss a lesson with AI teachers and AI classmates who use a whiteboard.
Download a lesson as editable .pptx slides or an interactive .html page.
Generate a classroom from a messaging app or an IDE through the OpenMAIC Skill.
Assign a model to each use of AI in openmaic.yml.

Getting started

The README says to clone the repository and run pnpm install. Then copy the example files with cp .env.example .env.local and cp openmaic.example.yml openmaic.yml, add a provider key, start a development database with pnpm db:up, uncomment the local DATABASE_URL line in .env.local, and run pnpm dev. For Docker, it describes copying .env.example to .env.local, adding your API keys and running docker compose up --build. It also describes a production build. The README has the exact commands. The README describes an optional configuration in ~/.openclaw/openclaw.json containing a hosted access code or self-hosted connection details. The README shows pnpm dependency installs, pip installs for the optional FunASR route, and Docker containers for the app, database and opt-in rendering service.

Limitations

The README labels v1.2.0-rc.1 a pre-release. The README lists Node.js 22.19+, pnpm 10+ and PostgreSQL 16 for source installation, and says DATABASE_URL is required. The README says 1.2.0 requires a long-running server and does not support Vercel or other serverless hosts. The README describes Compose as a personal, single-user installation listening on 127.0.0.1. The README warns that, without ACCESS_CODE, anyone who can reach the port can own, edit and delete the whole library. The README instructs network deployments to set a long random ACCESS_CODE in .env.local and, before the first start, a random PERSISTENCE_POSTGRES_PASSWORD through the shell or a .env file next to docker-compose.yml. The README warns that docker compose down -v deletes the volumes containing courses, generated media and runtime sessions. The README says ACCESS_CODE attempts are rate-limited only behind a trusted proxy with TRUST_PROXY_HEADERS=true. The README says MP4 export needs the opt-in render service and otherwise downloads a project ZIP. The README says the agent workbench is off by default, needs tool calling and must not set thinking.effort on its agent slot. The repository can change at any time; check its README before you install.

Data practices

The README says courses, folders, chat history, learner sessions and generated media live in PostgreSQL on the server, with media optionally in S3. The README says device settings, playback position and caches stay in the browser. The README says a stage ID allows reading a non-deleted course, while changes remain restricted to its owner. The README describes configured cloud and local AI providers, MinerU cloud or self-hosted document parsing, and AliDocMind cloud extraction or local ffmpeg/ffprobe media extraction. These choices determine which external services process submitted content. The README says a cloned voice clip is stored in the browser and sent to the configured VoxCPM backend on each synthesis.