Skip to content
OpenSourceSoftware.ai
Next AI Draw.io logo

Next AI Draw.io

DayuanJiang · Design

GitHub stars

36,490
—Awaiting first sync

Awaiting first sync

No measured star history yet

What is Next AI Draw.io?

Next AI Draw.io is for people who need diagrams but do not want to place every box and arrow by hand. You describe what you need and get a normal draw.io diagram that you can keep editing yourself or with the AI. It can run in a browser, as a desktop app, on your own server or inside AI coding tools.

Next AI Draw.io is an AI assistant built around the draw.io diagram editor. You describe a diagram in a sentence and the AI draws it on a real draw.io canvas, then you edit it by hand or ask the AI to change selected shapes. It runs as a web app, a desktop app for Windows, macOS and Linux, and an MCP server for AI coding tools.

Core capabilities

  • Draws flowcharts, architecture and sequence diagrams from a short description.
  • Turns each AI change into a version you can compare, restore or undo.
  • Works with 25 model providers, including local models through Ollama.

From a sentence to a draw.io diagram

The AI draws architecture diagrams, flowcharts, sequence diagrams and more on a draw.io canvas, using built-in AWS, Azure, GCP and Kubernetes icon libraries. Changes stream onto the canvas and the shapes the AI changed are highlighted. You can select shapes and ask about only those, and connectors can carry a flowing animation. Diagrams can have several pages and export as .drawio, .png, .svg or .drawio.svg.

  • Diagrams
  • draw.io
  • AI chat
  • Export

Model choice and the MCP server

Twenty-five providers are supported, including AWS Bedrock, OpenAI, Anthropic, Google AI, Azure OpenAI, Ollama, OpenRouter and DeepSeek, plus an OpenAI-compatible preset for other endpoints. In the web app you can enter your own API key, which stays in your browser. The project recommends a capable model, as small models tend to produce broken diagrams. The MCP server lets Claude Code, Codex, Cursor, VS Code and other clients draw diagrams with the same icon libraries, version history and export options, and its editor works offline.

  • Model providers
  • Ollama
  • MCP
  • Claude Code

Run it yourself under Apache-2.0

The public demo site has a usage limit that lifts when you add your own provider key. Desktop installers for Windows, macOS and Linux are on the GitHub Releases page. To self-host, run the ghcr.io/dayuanjiang/next-ai-draw-io Docker image, run it from source with npm, or deploy it to Vercel, Cloudflare Workers or Tencent EdgeOne Pages, and an offline guide covers intranet use. The project is released under the Apache-2.0 licence.

  • Self-hosted
  • Docker
  • Desktop
  • Apache-2.0

Where it fits

Use cases

  1. 01

    Draw a cloud architecture diagram

    Describe the system in the chat panel, such as the load balancer, services and database, and the AI draws it with the built-in AWS, Azure, GCP or Kubernetes icon libraries. Select shapes and ask the AI to change only those, or drag and restyle them by hand. Export the result as .drawio, .png, .svg or .drawio.svg.

  2. 02

    Turn a sketch or document into a diagram

    Upload a screenshot or a hand sketch and the AI redraws it as an editable diagram. You can also upload PDF, Markdown, code or other text files and get a diagram of their content. Every AI change appears as a version card with a thumbnail, so you can compare it with the canvas or restore an earlier one.

  3. 03

    Give a team access on your own server

    Run the Docker image from ghcr.io with your provider, model and API key set as environment variables. List several models in AI_MODEL or AI_MODELS_CONFIG so users do not need keys of their own. Set ADMIN_PASSWORD to manage models, access codes and quotas from the /admin page.