Claude Monet

Humans naturally think using visual structures — drawing diagrams, mapping concepts, and sketching incomplete ideas. However, most AI systems rely on text as the primary interaction medium. This project investigates how AI can participate directly in visual thinking by allowing users to draw or whiteboard ideas and receive meaningful AI feedback.

Our project is an interactive, AI powered whiteboard partner. We came up with this idea after thinking about the ways that AI is not used right now, and one way we thought of was as a partner to ideate with. The tool we build uses the white board canvas as an interface, communicating via the processing of drawing. Ways we shattered UI conventions:

• Interactive in real time, not just sending an image through a chat and receiving a response back.
• Voice integration as if you are bouncing ideas off of a companion or coworker.
• Visualizing ideas and getting feedback almost as if Claude is drawing next to you.
• Multi modal - uses Claude’s voice capabilities, tool-calling capabilities, vision capabilities.

The tools we used are:

• Forked this MIT licensed MCP Server and Agent skill that allows you to give Claude full canvas access: https://github.com/yctimlin/mcp_excalidraw
• Vite
• Redis for context management
• Node.JS, Typescript, ExpressJS
• Anthropic Claude SDK
• Responsive Voice.JS

None, it was all built from the start of the event.