interactive-edtech-mcp
MCP server suite that lets coding agents generate deep instructional content from external sources and operate classroom-interactive EdTech platforms including Padlet, Google Classroom, Kahoot!, Wayground, Wordwall, and Nearpod.
README
edu-agent-kit
An agent-powered knowledge base & content toolkit for educators. Build a local LLM-wiki for your teaching materials, generate deep instructional content from your own sources, and export/distribute it to Kahoot, Padlet, Nearpod, Wordwall, Wayground, Google Workspace, Google Classroom, and Firebase — from any MCP-capable agent.
🇹🇼 繁體中文說明:README.zh-TW.md | 詳細安裝:docs/INSTALL.md | API 申請:docs/API-SETUP.md
Works with Claude Code, Cursor, Codex, OpenCode, Antigravity, and any MCP client.
What it does
Native AI features in classroom tools produce shallow content, don't connect to agent workflows, and can't ingest your own materials. edu-agent-kit solves this end to end:
- Local LLM-wiki — scaffold a knowledge base for your teaching materials (templated folders, agent memory/
CLAUDE.md, dispatch commands), so an agent can orchestrate "material → knowledge base → generation → output." - Deep content generation — ingest files/URLs/web research, align to Taiwan's 108課綱, and generate quizzes/lessons/boards that are validated and Bloom-depth-scored.
- High-fidelity delivery — each platform gets its best supported path (official API where one exists, official import/export file otherwise) plus full Google Workspace + Classroom + Firebase output.
Four surfaces (cross-agent)
| Surface | What | For |
|---|---|---|
| MCP server | 44 tools | the universal layer — every agent |
CLI (edu-agent-kit) |
scaffold wiki, Google auth, exports, doctor | any terminal/agent |
| Skills | 5 educator workflows (繁中) | Claude & skill-capable agents |
| Claude Code Plugin | one-click bundle (MCP + skills) | the Claude ecosystem |
Platform support
| Platform | Create | Read |
|---|---|---|
| Padlet | ✅ official API (board + posts) | ✅ |
| Google Classroom | ✅ official API (assignments, materials, roster, grading) | ✅ |
| Google Docs / Slides / Forms / Sheets | ✅ official API | — |
| Google Drive | ✅ folders, upload, sharing | — |
| Firebase Hosting | ✅ publish material as a web page | — |
| Kahoot! | ⬇️ official .xlsx import |
✅ Reports API |
| Wayground (Quizizz) | ⬇️ official spreadsheet import | — |
| Wordwall | ⬇️ template content file | ✅ oEmbed |
| Nearpod | ⬇️ Google Slides .pptx |
— |
⬇️ = produces a file you upload in one step (no brittle browser automation; ToS-safe).
Quick start
# 1. Get the code & build
git clone https://github.com/FW1201/edu-agent-kit.git
cd edu-agent-kit
corepack enable pnpm
pnpm install && pnpm build
# 2. Scaffold your teaching knowledge base
node apps/cli/dist/index.js init --dir ~/my-teaching-wiki --template workflow \
--name "您的稱呼" --grades "國中" --subjects "數學" --agents claude --google docs,classroom
# 3. (optional) Authorize Google
node apps/cli/dist/index.js auth google --services docs,slides,forms,sheets,drive,classroom
# 4. Check setup
node apps/cli/dist/index.js doctor
Then connect your agent (below) and say things like "ingest raw/這課.pdf, generate a deep 8-question quiz for 國中, and build a Kahoot import file."
Tool catalog (44 tools)
- Content (6):
content_ingest_source,content_web_research,content_align_curriculum,content_generate_quiz/_lesson/_board - Knowledge base (3):
wiki_list_templates,wiki_scaffold,wiki_status - Padlet (5) · Google Classroom (13) · Google Workspace (8):
google_docs_create/_from_lesson,google_slides_create_from_lesson,google_forms_create_from_quiz,google_sheets_create,drive_create_folder/drive_upload_file/drive_set_sharing - Firebase (1):
firebase_deploy_hosting - Kahoot (3) · Wayground (1) · Wordwall (2) · Nearpod (1)
- Workflow (1):
workflow_generate_and_distribute(generate → deliver → distribute)
Connect your agent
See docs/INSTALL.md for full per-agent steps. Replace /ABS/PATH with your clone path.
Claude Code (plugin — recommended):
/plugin marketplace add FW1201/edu-agent-kit
/plugin install edu-agent-kit@edu-agent-kit
Or as a plain MCP server: claude mcp add edu-agent-kit -- node /ABS/PATH/edu-agent-kit/apps/server/dist/index.js
Cursor / Codex / OpenCode / Antigravity: add the MCP server (node /ABS/PATH/edu-agent-kit/apps/server/dist/index.js) to the agent's MCP config — exact JSON/TOML for each in docs/INSTALL.md.
Remote/HTTP: TRANSPORT=http PORT=3000 node apps/server/dist/index.js → http://127.0.0.1:3000/mcp.
Credentials
All optional — missing keys only disable their own tools. Where to get each: docs/API-SETUP.md.
| Variable | For |
|---|---|
GOOGLE_CLIENT_ID / GOOGLE_CLIENT_SECRET |
Google Docs/Slides/Forms/Sheets/Drive/Classroom |
PADLET_API_KEY |
Padlet |
KAHOOT_API_KEY |
Kahoot Reports API |
TAVILY_API_KEY |
web research |
FIREBASE_TOKEN / FIREBASE_PROJECT |
Firebase Hosting |
Google authorization is one-time: edu-agent-kit auth google (or node packages/adapters/google-classroom/dist/auth-cli.js).
Limitations & ToS
- No browser automation. Kahoot/Wayground/Wordwall/Nearpod "create" produces an official import/export file you upload in one step — a deliberate, ToS-safe choice.
- The MCP server has no LLM of its own:
content_generate_*returns a brief for your agent to author from, then validates + depth-scores the draft. - Google/Kahoot/Padlet/Firebase live features need their respective (sometimes paid) accounts.
- The 108課綱 alignment dataset is an extensible seed.
Architecture
edu-agent-kit/ pnpm monorepo
├── .claude-plugin/ plugin.json + marketplace.json
├── .mcp.json plugin MCP config
├── skills/ 5 educator skills (繁中)
├── packages/
│ ├── core/ sources/ mcp-shared/ models, ingestion, MCP utils
│ ├── wiki-kit/ LLM-wiki scaffolder
│ └── adapters/
│ ├── google-shared/ google-classroom/ google-workspace/ firebase/
│ └── padlet/ kahoot/ wayground/ wordwall/ nearpod/
└── apps/
├── server/ single MCP server (44 tools)
└── cli/ edu-agent-kit CLI
Development
pnpm build # build all packages
pnpm test # unit tests
pnpm typecheck # type-check
License
MIT © Kevin Wu (FW1201)
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。