AI-Powered Diagram Tool Launches on Next.js
A new open-source application called Next AI Draw.io combines the popular draw.io diagramming tool with large language models, enabling users to create, modify, and enhance diagrams through natural language commands. The project, hosted on GitHub at DayuanJiang/next-ai-draw-io, is built with Next.js and the Vercel AI SDK, supporting a wide range of AI providers including OpenAI, Anthropic, Google AI, DeepSeek, and ByteDance Doubao. The demo site is currently sponsored by ByteDance Doubao, which provides the underlying glm-4.7 model.
Key Features and Capabilities
The tool allows users to generate diagrams from scratch by describing them in plain language, upload existing diagrams or images for AI replication and enhancement, and extract content from PDF and text files to create diagrams. It displays the AI's reasoning process for supported models like OpenAI o1/o3, Gemini, and Claude, and maintains a comprehensive version history that lets users view and restore previous diagram versions. A real-time interactive chat interface lets users refine diagrams conversationally, while specialized support for cloud architecture diagrams covers AWS, GCP, and Azure. Animated connectors between elements add dynamic visualization.
Integration and Deployment
Next AI Draw.io can be used with AI agents such as Claude Desktop, Cursor, and VS Code via the Model Context Protocol (MCP). Users can run it locally after cloning the repository and installing dependencies with `npm install`, then starting the development server on port 6002. One-click deployment options include Tencent EdgeOne Pages, which provides a daily free quota for DeepSeek models, and Vercel. A native desktop app is available for Windows, macOS, and Linux, and the demo site can be tried directly without installation by bringing your own API key.
Provider and Model Support
The application supports a wide array of AI providers: ByteDance Doubao, AWS Bedrock (default), OpenAI, Anthropic, Google AI, Google Vertex AI, Azure OpenAI, Ollama, OpenRouter, AIHubMix, DeepSeek, SiliconFlow, ModelScope, SGLang, and Vercel AI Gateway. All providers except AWS Bedrock and OpenRouter support custom endpoints. For strong diagram generation, recommended models include Claude Sonnet 4.5, GPT-5.1, Gemini 3 Pro, and DeepSeek V3.2/R1, with Claude series noted for its training on draw.io diagrams with cloud architecture logos.
Administrative and Configuration Options
Administrators can configure multiple server-side models via the AI_MODELS_CONFIG environment variable or an ai-models.json file, making models available to all users without personal API keys. An admin panel, accessible by setting the ADMIN_PASSWORD environment variable and visiting /admin, allows managing models, access codes, features, observability, and quota through a web interface. The project is open source, with contributions and sponsorship welcomed to support the live demo site.