Wizard chess

• 

  Game screen

Inspiration

We wanted chess to feel like a conversation. Wizard Chess combines the atmosphere of magical chess with an AI agent that can understand and control the board directly. What it does

Wizard Chess is a 3D chess game controlled by voice, touch, mouse, or WebMCP. Players can learn through guided lessons, play with an AI mentor, or battle Stockfish at three difficulty levels.

How we built it

We used TypeScript, Vite, Three.js, chess.js, and Stockfish. The app exposes 29 WebMCP tools for moves, lessons, coaching, settings, saved games, and replays. Voice and WebMCP use the same game actions as the visible interface.

Challenges we ran into

The main challenge was keeping mouse, voice, and agent interactions synchronized with one game state. We also had to make the 3D board readable while handling voice timing, legal moves, lessons, matches, and replays.

Accomplishments that we're proud of

We built a complete chess experience where a person and an AI agent share the same board. It includes voice control, coaching, six lessons, Stockfish opponents, ratings, trophies, saved games, and replays.

What we learned

WebMCP works well for chess because the game has exact state and clear actions. Giving the agent structured board data is far more reliable than asking it to understand the position from pixels.

What's next for Wizard Chess

We want to add adaptive lessons, deeper post-game analysis, a hosted leaderboard, richer piece models, and more cinematic effects. The mentor could then remember recurring mistakes and recommend what to practise next.

Built With

• gemini
• kokoro
• openrouter
• vite
• webmcp