interactive-edtech-mcp

An MCP server suite that lets coding agents generate deep instructional content from external sources and operate classroom-interactive EdTech platforms — Padlet, Google Classroom, Kahoot!, Wayground (Quizizz), Wordwall, and Nearpod.

Works with any MCP-capable agent: Claude Code, Cursor, Codex, OpenCode, Antigravity, and more.

Why this exists

The native AI features in classroom-interactive tools tend to produce shallow content (clustered at the "remember/understand" Bloom levels, no rationales, no rubrics), are hard to wire into agent workflows, and can't easily ingest your own materials. This suite fixes that by splitting the problem in two:

1. A platform-agnostic depth layer — agents ingest your files/URLs/web research, align to 108課綱, and generate quizzes/lessons/boards that are validated and depth-scored (higher-order Bloom ratio, explanations, rubrics).

2. A high-fidelity delivery layer — each platform gets the best supported path: a real API where one exists, or an official import/export file where it doesn't.

Platform support matrix

⬇️ = produces a file you upload in one step. We deliberately do not use browser automation (brittle, ToS-risky).

Related MCP server: MCP Learning Server

Tool catalog (32 tools)

Content generation (platform-agnostic)

• content_ingest_source — parse a file (.pdf/.docx/.txt/.md/.csv/.json) or URL into structured material

• content_web_research — Tavily web search → source material

• content_align_curriculum — 108課綱 alignment scaffold (core competencies + objectives)

• content_generate_quiz / content_generate_lesson / content_generate_board — two-step brief → validate & depth-score

Padlet — padlet_get_board, padlet_list_posts, padlet_create_ai_board, padlet_add_post, padlet_seed_board

Google Classroom — classroom_list_courses, classroom_create_course, classroom_get_course, classroom_create_coursework, classroom_create_material, classroom_post_announcement, classroom_create_topic, classroom_list_students, classroom_add_student, classroom_list_teachers, classroom_list_submissions, classroom_get_submission, classroom_return_grade

Kahoot — kahoot_build_import_xlsx, kahoot_get_report, kahoot_analyze_results

Wayground — wayground_build_import_spreadsheet

Wordwall — wordwall_build_activity_content, wordwall_get_oembed

Nearpod — nearpod_build_slides_export

Workflow — workflow_generate_and_distribute (produce the platform artifact, then optionally create a Google Classroom assignment for it)

Installation

Prerequisites

• Node.js ≥ 18

• pnpm (corepack enable pnpm)

Build from source

git clone https://github.com/FW1201/interactive-edtech-mcp.git
cd interactive-edtech-mcp
pnpm install
pnpm build

The server entry point is apps/server/dist/index.js. Verify it:

node apps/server/dist/index.js   # prints: ...running on stdio (32 tools)

Environment variables

All keys are optional — tools whose credentials are missing simply return an actionable error; the rest keep working.

Copy .env.example to .env for local runs.

Connect your agent

Replace /ABS/PATH with the absolute path to your clone. The server speaks standard MCP over stdio, so any MCP client works.

Claude Code

claude mcp add interactive-edtech -- node /ABS/PATH/interactive-edtech-mcp/apps/server/dist/index.js

Or add to .mcp.json (project) / ~/.claude.json (global):

{
  "mcpServers": {
    "interactive-edtech": {
      "command": "node",
      "args": ["/ABS/PATH/interactive-edtech-mcp/apps/server/dist/index.js"],
      "env": { "PADLET_API_KEY": "...", "TAVILY_API_KEY": "..." }
    }
  }
}

Then in a session: > use content_generate_quiz to draft a deep 8-question quiz on 光合作用 for 國中一年級, then build a Kahoot import file.

Cursor

~/.cursor/mcp.json (global) or .cursor/mcp.json (project):

{
  "mcpServers": {
    "interactive-edtech": {
      "command": "node",
      "args": ["/ABS/PATH/interactive-edtech-mcp/apps/server/dist/index.js"],
      "env": { "PADLET_API_KEY": "..." }
    }
  }
}

Enable it under Settings → MCP, then ask the agent in Composer.

Codex

~/.codex/config.toml:

[mcp_servers.interactive-edtech]
command = "node"
args = ["/ABS/PATH/interactive-edtech-mcp/apps/server/dist/index.js"]

[mcp_servers.interactive-edtech.env]
PADLET_API_KEY = "..."
TAVILY_API_KEY = "..."

Run codex and the tools are available to the agent.

OpenCode

opencode.json (project) or ~/.config/opencode/opencode.json:

{
  "$schema": "https://opencode.ai/config.json",
  "mcp": {
    "interactive-edtech": {
      "type": "local",
      "command": ["node", "/ABS/PATH/interactive-edtech-mcp/apps/server/dist/index.js"],
      "environment": { "PADLET_API_KEY": "..." },
      "enabled": true
    }
  }
}

Antigravity

In Antigravity, open MCP settings (Agent panel → Manage MCP servers → Edit config) and add:

{
  "mcpServers": {
    "interactive-edtech": {
      "command": "node",
      "args": ["/ABS/PATH/interactive-edtech-mcp/apps/server/dist/index.js"],
      "env": { "PADLET_API_KEY": "..." }
    }
  }
}

Reload servers; the 32 tools appear to the Antigravity agent.

Remote / HTTP mode (optional)

TRANSPORT=http PORT=3000 node apps/server/dist/index.js
# Streamable HTTP endpoint at http://127.0.0.1:3000/mcp ; health at /health

Google Classroom OAuth (one-time)

# After setting GOOGLE_CLIENT_ID / GOOGLE_CLIENT_SECRET:
node packages/adapters/google-classroom/dist/auth-cli.js

It prints a consent URL, captures the redirect, and writes a token to GOOGLE_TOKEN_PATH. After that, the classroom_* and workflow_* Classroom features work.

Usage examples

1. Deep quiz → Kahoot

"Ingest syllabus.pdf, then generate a 10-question quiz on cell biology for 高中一年級 with rationales and rubrics, then build the Kahoot import workbook." The agent calls content_ingest_source → content_generate_quiz (brief, then validate) → kahoot_build_import_xlsx, and tells you the .xlsx path to upload.

2. Lesson → Nearpod

"Make an 8-slide interactive lesson on the water cycle with formative checks, then export it for Nearpod." → content_generate_lesson → nearpod_build_slides_export (.pptx to import via the Google Slides add-on).

3. End-to-end → Google Classroom

"Generate a discussion board on AI ethics, publish it to Padlet, and assign it in Classroom course 12345." → content_generate_board → workflow_generate_and_distribute (Padlet board live URL → classroom_create_coursework link assignment).

Limitations & Terms of Service

• No browser automation. For Kahoot/Wayground/Wordwall/Nearpod, "create" produces an official import/export file you upload in one manual step. This is 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 and depth-scores the draft. This keeps generation in your agent and enforcement in the server.

• Padlet's public API creates boards via the AI-board flow (no blank-board endpoint). Kahoot Reports API and Padlet API require paid plans.

• The 108課綱 alignment dataset is a representative seed, designed to be extended.

Architecture

interactive-edtech-mcp/            pnpm monorepo
├── packages/
│   ├── core/                      content models (Zod), pedagogy (Bloom/depth), validation
│   ├── sources/                   file/URL/web ingestion, 108課綱 alignment
│   ├── mcp-shared/                ToolDefinition contract, formatting, errors, http, auth
│   └── adapters/
│       ├── padlet/                official REST API
│       ├── google-classroom/      official API + OAuth2 (googleapis)
│       ├── kahoot/                .xlsx import + Reports API
│       ├── wayground/             spreadsheet import
│       ├── wordwall/              template content + oEmbed
│       └── nearpod/               Google Slides-compatible .pptx
└── apps/server/                   single MCP server (registers all tools) + workflow

Each adapter exports a ToolDefinition[] plus reusable functions; the server collects them and adds the content-generation and workflow tools.

Development

pnpm build        # build all packages (topological order)
pnpm test         # run all unit tests (44 tests)
pnpm typecheck    # type-check without emit
pnpm inspector    # launch the MCP Inspector against the server

Evaluations: evals/evaluation.xml (deterministic, offline).

License

MIT © Kevin Wu (FW1201)