Unlike image generators, this manipulates the actual draw.io XML. This enables a hybrid workflow: prompt the AI to build a base architecture, tweak it manually on the canvas, then ask the AI to refine it again.

Key features: - Cloud Ready: Native support for AWS, GCP, and Azure icon sets (not generic boxes). - Dynamic: Supports animated connectors and vector sketching (e.g., "draw a cat"). - Model Agnostic: Works with OpenAI, Anthropic, Bedrock, and Ollama.

I'd love feedback on the app, or ideas for v2 (e.g., add MCP feature). Any ideas are appreciated. Thanks!

Demo: https://next-ai-drawio.jiang.jp/ Repo: https://github.com/DayuanJiang/next-ai-draw-io