OpenMAIC Multi-Agent Interactive Classroom
One-click immersive multi-agent learning experiences for any topic or document.
- Source repo
- THU-MAIC/OpenMAIC
- Stars
- ★ 41k
- Last updated
- today
- License
- MIT
- Primary language
- TypeScript
- FA score
- 44/100 · Major gaps
At a glance
- How it runs
- Works with
- Universal · cross-platform
- Cost
- Free software; you pay for model usage
- Setup effort
- Medium · a few setup steps
- You'll need
- Typical use
- An instructor quickly converts a topic or slide deck into a full online course with slides, quizzes, and interactive exercises.
- Not a fit if
- Teams wanting zero-config use with no API keys or database
- Public multi-user deployments needing real auth and isolation
- Source review
- 44/100 · Major gaps 1 safety controls not found
What does this agent do, and when should you use it?
OpenMAIC is an open-source AI platform from Tsinghua University that turns any topic or document into a rich, interactive classroom experience. Powered by multi-agent orchestration, it generates slides, quizzes, interactive simulations, and project-based learning (PBL) activities, delivered by AI teachers and AI classmates who can speak, draw on a whiteboard, and engage in real-time discussions. Built with Next.js, React, and LangGraph, it supports major LLM providers like OpenAI, Anthropic, Gemini, and local Ollama, and integrates with OpenClaw to generate classrooms directly from messaging apps like Feishu and Slack. Licensed under MIT.
OpenMAIC ingests a user-provided topic or uploaded materials (documents, audio, video), and runs a two-stage generation pipeline (outline generation, then scene content generation) to build a complete course. It runs as a Next.js application, offering both the classic one-click generator and a new v1.0.0 Pro workbench with an agentic chat-based course builder. The platform supports multiple scene types, whiteboard drawing, text-to-speech (TTS), speech recognition (ASR), interactive HTML simulations, and project-based learning (PBL). It integrates with messaging apps via an OpenClaw skill, and can export courses as editable PowerPoint (.pptx) or self-contained interactive HTML, with optional PostgreSQL-backed persistence and Docker deployment.
- An instructor quickly converts a topic or slide deck into a full online course with slides, quizzes, and interactive exercises.
- A corporate trainer generates interactive training modules and distributes them directly in Feishu or Slack via OpenClaw.
- A self-learner describes a learning goal (e.g., "teach me Python from scratch in 30 min") and gets an immersive classroom with AI teachers and classmates.
- An edtech developer builds a customized learning system on top of OpenMAIC, leveraging its pluggable storage, model, and service providers.
- A school or institution needing offline or intranet courses uses the classroom ZIP export to play content in an air-gapped environment.
How do you install or deploy this agent?
- Clone the repo: git clone https://github.com/THU-MAIC/OpenMAIC.git 2. Install dependencies: cd OpenMAIC && pnpm install 3. Copy environment: cp .env.example .env.local 4. Configure at least one LLM provider key (e.g., OPENAI_API_KEY) 5. Run locally: pnpm dev, then open http://localhost:3000. Docker deployment is also supported: docker compose up --build. Optional: enable PostgreSQL persistence and the Pro workbench.
How do you use this agent?
Open http://localhost:3000, enter a topic or upload materials on the home page, and click generate to build a classroom. Watch AI teachers lecture, participate in quizzes, PBL activities, or discussions with AI classmates. Configure TTS/ASR and other providers in Settings. With OpenClaw integration, you can generate classrooms directly from a chat app.
What are this agent's strengths and limitations?
- One-click generation of complete courses with multiple interactive scenes, drastically reducing course creation effort.
- Multi-agent classroom provides realistic student-teacher interaction with whiteboard, voice, and real-time discussion.
- Model-agnostic: supports major LLM providers, local models (Ollama), and multiple TTS/ASR services.
- Rich export formats (PPTX, HTML, ZIP) enable offline use and content reuse.
- MIT license allows commercial use and modification.
- Generation quality heavily depends on the configured LLM provider; different models yield varying results.
- Requires at least one LLM API key, incurring potential API costs.
- Advanced features like video export and advanced document parsing need additional service components (Docker containers, external APIs).
- Pro workbench and server persistence require extra PostgreSQL setup, which may be a hurdle for non-technical users.
- As a relatively new project, updates are rapid and some documentation may lag.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| OpenMAIC Multi-Agent Interactive Classroom This agent | 44 · Major gaps | Web appFree + model costs | ★ 41k | today | TypeScript | — |
| Presenton – Open-Source AI Presentation Generator & API | 56 · Major gaps | Self-hosted serviceFreemium | ★ 11k | today | TypeScript | — |
| CraftBot | 40 · Major gaps | CLIFree + model costs | ★ 396 | 13d ago | Python | OpenAI API · Claude API |
| Team9 - Collaborative Workspace for AI Agents | 9 · Major gaps | Hosted serviceFreemium | ★ 2k | 4mo ago | TypeScript | — |
How does FollowAgents rate this agent?
Why each dimension lost points
Evidence shows: repository includes a security policy (SECURITY.md) with vulnerability reporting; OpenClaw skill claims confirmation at each step; SSRF hardening mentioned; many dependencies but no security audit evidence; external effects (web search, export) described but permission boundaries not detailed; rollback mechanism not mentioned; source attribution via license and third-party components. Deductions: least privilege not concretely implemented, user confirmation only claimed, data flow transparency insufficient, sensitive data handling details missing, dependency security lacks audit or vulnerability scan evidence, external effects permissions unclear, rollback missing, source attribution partial.
Evidence shows: README consistent with code structure, CI includes lint, typecheck, unit and e2e tests, indicating good self-consistency; many dependencies but no availability guarantee; failure messages tested in some cases (e.g., CORS failure returns error code), but overall failure handling documentation insufficient. Deductions: self-consistency supported by tests but not fully verified; dependency availability lacks lockfile or mirror backup; failure messages only partially tested, not covering all critical paths.
Evidence shows: README clearly defines target users (learners, teachers) and use cases (classroom generation, interactive learning); capability boundaries defined via feature list and architecture; trigger precision in OpenClaw skill has confirmation steps but not detailed trigger conditions; environment fit supports multiple deployments (local, Docker, Vercel) and many LLM providers. Deductions: trigger precision only partially described, not all trigger conditions; environment fit broad but lacks detailed configuration guide.
Evidence shows: README well-structured with quick start, features, use cases; install notes detailed (requirements, config, start); naming stability has version numbers but no naming convention; examples and FAQ present; known limitations not explicitly listed; license MIT with third-party components noted; changelog exists; maintenance responsibility via contribution guide and contact. Deductions: naming stability not explicitly stated; known limitations not separately listed; maintenance responsibility partial (contact but no clear maintenance team).
Evidence shows: output usability high (multiple export formats: PPTX, HTML, ZIP); marginal value clear (one-click interactive classroom, multi-agent interaction); cost-benefit not detailed (no performance benchmarks or cost estimates). Deductions: cost-benefit lacks data support.
Evidence shows: feature claims partially supported by tests (e.g., e2e); cross-source corroboration limited (only internal tests); fact-inference separation unclear. Deductions: claim traceability insufficient, no specific test mapping; cross-source corroboration lacks external validation; fact-inference separation not explicit.
- Not found in source: rollback or recovery pathBack up first, or work on a git branch or snapshot, so its changes can be undone.
- Many dependencies without security audit or vulnerability scan evidence; recommend dependency security review before deployment.
- OpenClaw skill claims confirmation but lacks implementation details; verify actual behavior.
- Rollback mechanism not mentioned; errors may be irreversible.
- Sensitive data handling details insufficient, especially API keys and user data storage.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
v1.2.1 OpenMAIC v1.2.1 — Web research: Kimi, DeepSeek and Firecrawl
Release notes excerpt
1.2.1 is a regular patch release. Its main theme is web research: Kimi, DeepSeek and Firecrawl join as web-search providers, and a new webFetch slot lets the agent's fetch_url tool use a hosted reader instead of the built-in one. It also contains one breaking change, the removal of the legacy director-graph classroom chat ([#1913](https://github.com/THU-MAIC/OpenMAIC/pull/1913)), and removes retired models from the built-in catalogs ([#1939](https://github.com/THU-MAIC/OpenMAIC/pull/1939)). Read Breaking Changes before upgrading from 1.2.0. Breaking Changes …View on GitHub ↗ -
v1.2.0 OpenMAIC v1.2.0 — Server-first, with classroom export
Release notes excerpt
Up to 1.1.x the browser drove course generation one request per step and kept the courses, the provider keys and the model settings itself. Closing the tab stopped a course halfway, every browser had to be set up on its own, a deployment could offer providers but not fix which models its users get, and the headless API ran a separate pipeline that drifted from the web app. 1.2.0 moves all of this to the server: courses and media, identity, model configuration and course generation live there, and the browser is a client. Classrooms can now also leave OpenMAIC: as a PPTX that keeps every scene, …View on GitHub ↗ -
v1.1.4 OpenMAIC v1.1.4 — Security
Release notes excerpt
A security release. Providers that use server credentials only send models the deployment authorizes, render jobs are bound to the requesting owner, and deleted courses stop serving their legacy media. Read Breaking change before upgrading, and upgrade the render-service together with or before the app. The same fixes ship in [v1.2.0-rc.2](https://github.com/THU-MAIC/OpenMAIC/releases/tag/v1.2.0-rc.2). Thanks to @13ob0 and @Duang777 for the reports. Advisories …View on GitHub ↗