Generative Learning

Generate the learning. Not the canvas.

We build the learning surface once. Your AI agent personalizes the lesson endlessly.

You learn on a chessboard or triangle canvas built for the subject. Your AI agent generates the lesson inside that surface. It does not invent a throwaway canvas, and personalizes pace, explanations, and practice in real time.

Live demo

Teaching tools work in Codex on ChatGPT desktop, or in Chrome with chrome://flags/#enable-webmcp-testing enabled.

Working surfaces

Chess

A persistent board, coach, and chess tools. Open /chess.

Triangles

GAN constructions, a figure canvas, and triangle tools. Open /triangles.

How it works

Each surface registers subject tools on document.modelContext (WebMCP). Your AI agent does not build the interface. It creates the lesson inside it, then guides you through it.

Home  →  open-page(chess | triangles)
             ↓
      subject surface + tools
             ↓
      your AI agent teaches on that surface

Chess tools

get-board-state, make-move, get-possible-moves, set-position, annotate-board, create-lesson, add-lesson-step, enter-learn-mode, and others on the chess page.

Triangle tools

get-figure-state, apply-gan, set-figure, move-point, rotate-figure, mark-figure, measure-figure, create-lesson, and others on the triangles page.

Home-page tools (list-pages, open-page) only navigate. Teaching tools appear after that page loads.

Getting started

git clone https://github.com/matipojo/WebMCP-Generative-Learning
cd WebMCP-Generative-Learning
npm install
npm start

Open http://localhost:3000 (the home page), then Chess or Triangles.

Credit

This chess game is based on React-Chess by @szabolcsthedeveloper.