What problem does it solve? AI chat interfaces normally return plain text, making it hard to present visual architecture diagrams, flowcharts, or whiteboard-style sketches. This Skill teaches agents how to build MCP servers that return interactive Excalidraw diagrams using the MCP Apps extension. ## Core Features & Use Cases - Interactive Diagram Rendering: Return HTML resources that render Excalidraw canvases directly in Claude Desktop, ChatGPT, VS Code, and Goose. - Streaming Updates: Progressively build diagrams by sending multiple tool responses that share the same resource URI, with viewport control over scroll position and zoom. - Flexible Deployment: Run as a Vercel serverless function, a Docker container, or a local Node.js process over stdio. - Use Case: Build an MCP tool that takes a list of system components and generates a hand-drawn architecture diagram with rectangles, labels, and bound arrows that users can open fullscreen and edit. ## Quick Start Ask the agent to create an Excalidraw diagram in your MCP server showing a simple client-server-database architecture.