- 画架构图/流程图时不想手拖：内置 AWS/Azure/GCP/K8s 图标库，让模型直接产出带正确图标的 draw.io XML，比从零手摆省时间；换通用绘图工具就得自己找图标、对坐标
- 已有图要局部改：选中形状后只让 AI 改这部分，改动流式上屏并高亮，比整图重生成可控；换纯文本生成方案容易把没让改的部分一起改坏
- 在 Claude Code / Cursor 里写代码时顺手出图：走 MCP server 让 agent 直接产出图，不用切到浏览器；没有 MCP 的话只能自己复制 XML 再粘进 draw.io
- 把截图、草图或 PDF/代码重绘成图：适合把已有材料转成可编辑的 draw.io 文件，而不是只拿到一张位图
next-ai-draw-io
DayuanJiangApache-2.0活跃维护 有专家点评用聊天方式让 AI 直接绘制和编辑 draw.io 图表，提供网页版、桌面版和 MCP 服务器，可接入 Claude Code、Cursor 等编码工具。
Draw and edit draw.io diagrams by chatting with an AI. Web app, desktop app, and an MCP server for Claude Code, Cursor, Codex and VS Code.
v0.5.1发布于 2026-10-31 02:25 UTC36,873star +91 / 7 天3,844fork同步于 2026-11-11 05:20 UTC
专家点评
上手难度：进阶更新于 2026-11-11在真实 draw.io 画布上用对话改图，把手工拖拽换成自然语言编辑
适用场景
踩坑
- 模型能力不够时容易吐出格式错误的 draw.io XML，README 自己就提示小模型不行；用本地 Ollama 或便宜小模型时大概率撞上，issue 里 edit_diagram failed 这类报错被反复提出
- 必须先配 .env.local 里的模型提供商和 API Key 才能出图，clone 完直接 npm run dev 是空转；默认端口是 6002 不是 3000，按习惯访问会打不开
- 版本卡片、对比恢复这套机制意味着 AI 改动是分步落盘的，多轮连续编辑后想精确回退到某一版要认这套卡片，而不是只靠编辑器撤销栈
同类对比
如果只是要 LLM 产出 Mermaid/PlantUML 文本图，用不着它，直接让模型写代码块即可；要的是 draw.io 原生可编辑文件、且接受自带 API Key 调云端模型，才轮到它
git clone https://github.com/DayuanJiang/next-ai-draw-io
cd next-ai-draw-io
npm install
cp env.example .env.local
npm run dev 硬件要求按 README 说明
README 未说明显存或 GPU 要求；网页版和桌面版均通过 API 调用云端模型，也可通过 Ollama 使用本地模型。
项目说明
由 README 自动翻译提炼用聊天在真实 draw.io 画布上生成和编辑图表，支持网页、桌面和 MCP 三种用法
用聊天方式让 AI 直接绘制和编辑 draw.io 图表，提供网页版、桌面版和 MCP 服务器，可接入 Claude Code、Cursor 等编码工具。
核心能力
- 一句话生成架构图、流程图、时序图，内置 AWS/Azure/GCP/K8s 图标库
- 选中形状后让 AI 只改这些形状，改动流式上屏并高亮
- 每次 AI 修改生成版本卡片，可对比、恢复或一键 Ctrl+Z 撤销
- 支持上传截图、草图或 PDF/代码等文本文件，由 AI 重绘成图
- 提供 MCP 服务器，可在 Claude Code、Cursor、VS Code 中直接画图
容易踩的坑
需先在 .env.local 中填入模型提供商和 API Key，否则无法出图；默认开发端口是 6002，不是常见的 3000；小模型容易生成格式错误的 draw.io XML，建议选能力较强的模型。
内容由模型根据 README 生成，可能不完整或不准确，以官方文档为准。看原文 README
- v0.5.1最新正式版2026-10-31 · 2026-10-31 02:25 UTC
What's New in v0.5.1
Three weeks of fixes for the desktop app: the AI works on the page you are viewing, a style button sits next to the selection, PDF export keeps the crop size on Windows, every answer shows its tokens and time, and models can think or not by a switch. It also brings the security updates of Electron 42.
New Features
- Style button next to the selection: a brush button next to the selected shapes opens draw.io's format panel, for changing colors, lines and text by hand. With the chat panel hidden, a sparkle button next to it opens the panel to ask the AI (#1039)
- Tokens, time and estimated cost under each answer: hover the numbers for the details, such as how much input was read from the cache (#1016), and an estimated cost priced from models.dev (#762 by @Zhangxiaoyi688)
- Thinking switch: models whose API can turn thinking on or off get a switch in the model settings (#911 by @lq-kiren)
- Named styles: the model defines a style once and draws with it by name, instead of repeating the same style string on every shape (#969)
- Models that cannot read images are marked in the settings, with a warning when an image is attached to one (#1018)
- MCP server card on the start screen, with a short video, and a docs button (#990, #991, #1010)
Improvements
- The AI reads and draws the page you are viewing in a multi-page diagram, instead of always the first page (#974)
- The canvas goes with a message only when the model needs it: earlier messages stay unchanged, so prompt caching works, and hand edits made between answers are no longer turned back (#1024)
- Saved prompts fill the message box instead of sending at once, so they can be changed first (#1036)
- Pictures in the diagram reach the model as short references instead of thousands of characters of image data (#1007)
- Time so far while a model draws without streaming (#1011)
- Fixes from a user simulation: the "changed by AI" outline fades after a few seconds, Enter sends, saved diagrams are one click away, PNG export is 2x on white with a margin, and each edit's reply says what changed (#996)
- New app icon (#985)
- MiniMax: current text models suggested (#1026 by @octo-patch)
Bug Fixes
- PDF export on the desktop: Export in the PDF dialog saves the PDF directly with the diagram's own page size, so Crop no longer adds large margins and wide diagrams are no longer turned 90 degrees on Windows (#1038)
- Shapes stay on their layer when the AI edits a page whose other layer comes first in the file (#1037)
- Custom data on shapes is kept on their
<object>wrapper (#1017) - Retries, stops and cut-off answers: a retry after a tool error sends the current canvas, Stop also ends the screenshot check and the connection test, and an answer that ran out of output tokens before drawing says so with a Retry button (#987)
- The daily limit message no longer crashes the page (#1006)
Security
- Credentials: the Test and model list endpoints refuse credentials that are not text (#1021)
- Dependencies: Electron 42 (#1029, after Electron 41.10.7 in #973),
@modelcontextprotocol/sdk1.32.0 (#1028),@opennextjs/cloudflare1.20.2 (#1000),proxy-addr,shell-quoteandwrangler(#971)
MCP Server (0.4.1 to 0.4.7)
- The preview is the web app's canvas: version cards, compare and restore, and
get_selectionfor "this box" (#984) - Saved sessions,
restore_versionand the history on disk (#981); a screenshot attached tocreate_new_diagramandedit_diagramresults on request (#983) load_diagramopens.drawio.svgfiles (#982) and accepts inlinexml(#946 by @caoxiaole07); a summary of the user's hand changes, an XML reference by topic and preview options (#982); drawing guide hints and your own drawing rules (#980)- Install: one-click install for VS Code and Cursor, and a Gemini CLI extension (#1015); the Windows preview opens withou
- v0.5.02026-10-10 · 2026-10-10 03:06 UTC
查看更新说明
What's New in v0.5.0
The first release since May. The web UI is rebuilt around the canvas, draw.io is bundled with the app, the settings dialog is redone, and the MCP server reaches feature parity with the web app. Desktop users get every fix that landed on
mainover the last five months.New Features
- Canvas-first UI with a bundled draw.io: the canvas fills the window and the chat floats on its right. draw.io v32.0.2 is served from the app's own origin (no separate draw.io container, works offline), so the app drives the editor directly: every AI change is a version card with a thumbnail that can be compared, restored or undone; an AI change is one Ctrl+Z step; changed shapes are highlighted; selected shapes are sent to the model; draw.io's own toolbar, menus (auto layout, Mermaid, CSV import) and page tabs are available. New start screen, session rename and switcher, and opening
.drawiofiles (#957, #962) - Settings dialog rebuilt: tabs for Models, General, Drawing and About. Adding a provider is three numbered steps (connection, models, test), with a searchable provider picker, one status line per provider, Bedrock sign-in by API key or access keys with a session token, and a model picker that lists every model as working, untested or failed (#958)
- MCP server 0.3.0, feature parity with the web app: the drawing guide and 30 icon libraries over MCP,
screenshot_diagramfor visual checks, all-or-nothingedit_diagram, sessions auto-saved to.drawiofiles (DRAWIO_DATA_DIR), preview themes and dark mode, editable SVG export. The web app now reuses the MCP server's diagram code (#951) - Multi-page documents in the MCP server:
add_page, edit a page by id, name or index, export specific pages (#862) load_diagramMCP tool: load a.drawiofile from disk into the session, including draw.io's compressed format (#893)- Admin panel at
/admin: manage server settings in a web UI instead of environment variables, enabled by settingADMIN_PASSWORD; secrets are masked, serverless hosts get a read-only view (#866). Its copy is rewritten in plain words in all four languages (#963) - Model lists and catalog: 18 providers can fetch their model list into a searchable picker; a models.dev snapshot flags models without tool calls, which cannot draw; the Test button checks all models at once and shows response times (#951)
- New providers: AIHubMix (#865), Atlas Cloud (#896) and MiMo by Xiaomi (#887); MiniMax default model upgraded to M3 (#857)
- Comma-separated
AI_MODEL:AI_MODEL=a,b,cis enough for a quick multi-model server setup (#870) - Quota by the CDN's client IP:
CLIENT_IP_HEADERandORIGIN_SECRETstop quota bypass through a fakedX-Forwarded-Forheader (#956)
Improvements
- Claude 4.7 and later, GPT 6: requests retry without sampling settings and with adaptive thinking on models that reject
temperature; GPT 6 models show their reasoning; suggested models updated to Claude 5 and GPT 6; Bedrock model ids fixed to use inference profiles (#951) - Provider error hints: provider errors are classified into 12 kinds, each with a hint the user can act on and an "Open model settings" button (#951). When the server's own key is refused, the chat shows a quota hint instead of two unhelpful lines (#953)
- Output token budget:
maxOutputTokensis always sent (#915) and the default is raised to 64000 with a retry at the provider's limit, so reasoning models reach the tool call instead of ending without a diagram (#927) - Self-hosted deployments: the MCP server card is hidden when
NEXT_PUBLIC_SELFHOSTEDis true (#711);ANTHROPIC_AUTH_TOKENis accepted as an alternative toANTHROPIC_API_KEY(#853); suggested models refreshed for all providers (#863);env.examplelists every provider (#881) - Accessibility: the viewport no longer blocks pinch zoom on phones (#881)
Security
- SSRF in
/api/parse-url: private and internal URLs ar
- Canvas-first UI with a bundled draw.io: the canvas fills the window and the chat floats on its right. draw.io v32.0.2 is served from the app's own origin (no separate draw.io container, works offline), so the app drives the editor directly: every AI change is a version card with a thumbnail that can be compared, restored or undone; an AI change is one Ctrl+Z step; changed shapes are highlighted; selected shapes are sent to the model; draw.io's own toolbar, menus (auto layout, Mermaid, CSV import) and page tabs are available. New start screen, session rename and switcher, and opening
- v0.4.160.4.162026-05-21 · 2026-05-21 00:31 UTC
查看更新说明
What's New in v0.4.16
Bug Fixes
- Electron Window Close: Override draw.io iframe
beforeunloadso the window can be closed after editing a diagram on macOS/Windows (fixes #815) (#840) - Kimi Reasoning Models: Use
createDeepSeekfor the Kimi provider soreasoning_contentround-trips correctly in multi-turn conversations (fixes #824) (#825) - Access Code Config: Always re-fetch access code config when the settings dialog opens, so the input field shows up after server-side
ACCESS_CODE_LISTchanges (fixes #811) (#816) - Model Config Sidebar: Add
min-h-0to the AI Model Configuration sidebar so the "Add model" button stays reachable when many models are added (fixes #576) (#818)
Improvements
- All Draw.io Themes: Replace the theme toggle in the settings panel with a dropdown exposing all 6 draw.io themes — kennedy, atlas, dark, min, sketch, simple (closes #499) (#835)
Developer Experience
- Dead Code Cleanup: Remove renderer-side workarounds and IDB recovery code left over from earlier #815 fix attempts now that #840 is the real fix (#841)
- E2E Test Fix: Resolve iframe toolbar strict-mode violation in the
can interact with draw.io toolbartest (#837) - Biome Schema Alignment: Bump
biome.json$schemato2.4.14so the repo matches the version installed by the Auto Format workflow (#832) - Dependency Updates: Security update for
nextto v16.2.6 (#842), security update forelectronto v39.8.5 (#789), security update for@xmldom/xmldomto v0.9.10 (#821), core framework packages major update (#724), and other minor/patch dependency updates (#829)
Contributors
- @DayuanJiang
- @LaaraibAhmed
- @octo-patch
- @ruguoba
Downloads
Platform File Notes macOS (Intel) Next-AI-Draw.io-0.4.16.dmg macOS (Apple Silicon) Next-AI-Draw.io-0.4.16-arm64.dmg Windows (Installer) Next.AI.Draw.io.Setup.0.4.16.exe Recommended - installs to Program Files, auto-updates Windows (Portable) Next.AI.Draw.io.0.4.16.exe No installation needed, just run Linux (AppImage) Next-AI-Draw.io-0.4.16.AppImage Linux (deb) next-ai-draw-io_0.4.16_amd64.deb Linux (rpm) next-ai-draw-io-0.4.16.x86_64.rpm ⚠️ Note: The app is not code-signed on macOS. You may see security warnings:
- macOS: "App is damaged" → Right-click the app → Open → Open, or run:
sudo xattr -rd com.apple.quarantine "/Applications/Next AI Draw.io.app" - Windows: Now code-signed with SignPath!
Full Changelog: https://github.com/DayuanJiang/next-ai-draw-io/compare/v0.4.15...v0.4.16
- Electron Window Close: Override draw.io iframe
- v0.4.150.4.152026-04-14 · 2026-04-14 15:01 UTC
查看更新说明
What's New in v0.4.15
New Features
- Personal My Templates Library: Added personal My Templates library alongside Quick Examples for saving and reusing diagram templates (#773)
- XML+SVG Export: Added xmlsvg export option for diagrams (#761)
- RPM Package for Linux: Added rpm target to electron-builder for Linux distribution (#806)
Improvements
- Tool Demo Updates: Updated tool demos on the English about page (#598)
- MCP Tool Description: Improved create_new_diagram tool description to prevent misuse (#803)
- Chat Lobby Panels: Show all enabled panels on chat lobby instead of only templates (#802)
Bug Fixes
- QvQ Vision Model Support: Allow QvQ (Qwen Visual QA) models to use image input (#808)
- Qwen3.5 Vision Model Support: Allow Qwen3.5 models to use image input (#800)
- CPU Usage Optimization: Reduce CPU usage during large XML streaming (#801)
- Undo/Redo Restoration: Remove redundant status call after autosave to restore undo/redo (#780)
- MCP Security - Localhost Binding: Bind MCP server HTTP to 127.0.0.1 only (#787)
- MCP Security - Body Size Limit: Add 10MB body size limit to MCP HTTP endpoints (#791)
- Zoom & IndexedDB Fix: Fix zoom reset on drag and IndexedDB version conflict (#776)
- System Message Merging: Merge system messages for custom OpenAI-compatible endpoints (#774)
- Dependency Updates: Update minor and patch dependencies (#767)
Developer Experience
- Remove MCP Preview Labels: Cleaned up MCP preview labels from the UI (#790)
- Electron Build Improvement: Always bundle latest draw.io version in Electron builds (#792)
- Contributing Guidelines: Added AI tools guidelines to CONTRIBUTING.md (#771)
Contributors
- @astordu
- @DayuanJiang
- @khanhthanhdev
- @octo-patch
- @yuzhichang
- @zongxi1115
Downloads
Platform File Notes macOS (Intel) Next-AI-Draw.io-0.4.15.dmg macOS (Apple Silicon) Next-AI-Draw.io-0.4.15-arm64.dmg Windows (Installer) Next.AI.Draw.io.Setup.0.4.15.exe Recommended - installs to Program Files, auto-updates Windows (Portable) Next.AI.Draw.io.0.4.15.exe No installation needed, just run Linux (AppImage) Next-AI-Draw.io-0.4.15.AppImage Linux (deb) next-ai-draw-io_0.4.15_amd64.deb Linux (rpm) next-ai-draw-io-0.4.15.x86_64.rpm New! ⚠️ Note: The app is not code-signed on macOS. You may see security warnings:
- macOS: "App is damaged" → Right-click the app → Open → Open, or run:
sudo xattr -rd com.apple.quarantine "/Applications/Next AI Draw.io.app" - Windows: Now code-signed with SignPath!
Full Changelog: https://github.com/DayuanJiang/next-ai-draw-io/compare/v0.4.14...v0.4.15
- v0.4.140.4.142026-03-30 · 2026-03-30 12:39 UTC
查看更新说明
What's New in v0.4.14
New Features
- Novita AI Provider: Added Novita AI as a new LLM provider with OpenAI-compatible API and open-source model support (#758)
Improvements
- GLM Vision Model Check: Added checks for GLM text and visual model naming conventions (#741)
- MiniMax M2.7 Model: Upgraded MiniMax default model to M2.7 with high-speed variant (#737)
Bug Fixes
- MCP Server CORS Security: Restricted CORS to same-origin only, preventing external sites from accessing the MCP server (#757)
- Settings Dialog Overflow: Fixed settings dialog overflowing outside the viewport with thin scrollbar styling (#743)
- Electron Startup on Windows/Linux: Resolved Electron startup failures caused by port conflicts on Windows and Linux (#709)
- IndexedDB Recovery: Fixed IDB closing recovery and restored diagram-only sessions after refresh (#648)
Developer Experience
- CI Cleanup: Removed
npm auditstep from CI that was blocking unrelated PRs (#744)
Contributors
- @Alex-wuhu
- @Biki-dev
- @DayuanJiang
- @octo-patch
- @sbilly
Downloads
Platform File Notes macOS (Intel) Next-AI-Draw.io-0.4.14.dmg macOS (Apple Silicon) Next-AI-Draw.io-0.4.14-arm64.dmg Windows (Installer) Next.AI.Draw.io.Setup.0.4.14.exe Recommended - installs to Program Files, auto-updates Windows (Portable) Next.AI.Draw.io.0.4.14.exe No installation needed, just run Linux (AppImage) Next-AI-Draw.io-0.4.14.AppImage Linux (deb) next-ai-draw-io_0.4.14_amd64.deb ⚠️ Note: The app is not code-signed on macOS. You may see security warnings:
- macOS: "App is damaged" → Right-click the app → Open → Open, or run:
sudo xattr -rd com.apple.quarantine "/Applications/Next AI Draw.io.app" - Windows: Now code-signed with SignPath!
Full Changelog: https://github.com/DayuanJiang/next-ai-draw-io/compare/v0.4.13...v0.4.14
用户点评
登录用户的真实使用记录，可按版本和硬件筛选共 0 条
还没有人点评。第一个说出真实体验的人最有价值。
Next AI Draw.io
Draw and edit draw.io diagrams by chatting with an AI.
Live Demo · Docs · Gallery · Desktop App · MCP Server
https://github.com/user-attachments/assets/66b9f12f-219f-4d62-acc0-0725e6850eec
Describe a diagram in a sentence and the AI draws it on a real draw.io canvas. Edit it by hand like any draw.io file, or select a few shapes and tell the AI what to change. Every AI change is a version you can compare, restore or undo, and the result exports as .drawio, .png or .svg.
Available as a web app, a desktop app for Windows, macOS and Linux, and an MCP server for AI agents such as Claude Code, Cursor and VS Code.
Highlights
Draw
- Architecture diagrams, flowcharts, sequence diagrams and more from one sentence, with built-in AWS, Azure, GCP and Kubernetes icon libraries
- Connectors can carry a flowing animation
- Upload a screenshot or a sketch and the AI redraws it; upload PDF, Markdown, code and other text files and get a diagram of their content
Edit
- Edit by chat: changes stream onto the canvas and the shapes the AI just changed are highlighted
- Ask about a selection: select shapes on the canvas and the AI changes only those
- Versions and undo: every AI change is a version card with a thumbnail; compare it with the canvas, restore it or undo it, and Ctrl+Z on the canvas also takes back an AI change in one step
- It is a normal draw.io diagram: double-click to rename, drag, restyle, use several pages, and export as
.drawio,.png,.svgor.drawio.svgat any time
Use
- 25 model providers, including an OpenAI-compatible preset for any other URL; enter your own API key in the browser and it stays on your machine
- Models that reason show their thinking
- Dark mode; the interface is available in English, Simplified Chinese, Traditional Chinese and Japanese
Examples
More examples, each with its prompt and model, are in the gallery.
Use it
Online demo
Open next-ai-drawio.jiang.jp, nothing to install. The demo has a usage limit; click the settings icon in the chat panel and enter your own provider and API key to lift it. The key stays in your browser and is never sent to the server for storage.
Desktop app
Download the Windows, macOS or Linux installer from the Releases page.
In your AI agent (MCP)
https://github.com/user-attachments/assets/6f53a67b-ab21-4113-9fee-7fb32f9de678
Through MCP (Model Context Protocol, the protocol AI agents use to call outside tools), Claude Desktop, Cursor, VS Code and others can draw draw.io diagrams directly. Add this to your client's MCP configuration:
{
"mcpServers": {
"drawio": {
"command": "npx",
"args": ["@next-ai-drawio/mcp-server@latest"]
}
}
}
In VS Code or Cursor, one click adds it: 
For Claude Code, one command does it (--scope user makes it available in every project):
claude mcp add --scope user drawio -- npx @next-ai-drawio/mcp-server@latest
It is also a plugin in the Claude Code community marketplace (Claude Code 2.1.275 or later):
/plugin install next-ai-drawio --marketplace anthropics/claude-plugins-community
Then ask the AI for "a flowchart of user authentication with login, MFA and session management" and the diagram appears in your browser as it is drawn. The MCP server has most of the web app's drawing features:
- The same drawing rules, icon libraries (AWS, Azure, GCP, Kubernetes and more) and canvas as the web app: the AI's changes are outlined, one Ctrl+Z takes them back, and the AI reads the shapes you select
- A screenshot tool, so the AI looks at the rendered diagram and fixes it
- Version history, several pages, export as
.drawio,.png,.svgor.drawio.svg, and auto-save to~/.next-ai-drawio/; the draw.io editor ships inside the package, so the preview works offline
The MCP server page has the setup for Codex, VS Code, Cursor and other clients, every tool, and how it compares with other draw.io MCP servers.
Compared with draw.io's built-in AI
draw.io has its own Generate tool (Arrange > Insert > Generate). It runs inside draw.io on a free hosted model or your own API key, and it can change a diagram, page or selection that you attach to its chat. Next AI Draw.io sends the current canvas, including your hand edits, with every message, turns every AI change into a version you can compare or restore, and works with 25 providers, including local models through Ollama, on the demo site, in the desktop app or on your own server.
Self-host
Run locally
git clone https://github.com/DayuanJiang/next-ai-draw-io
cd next-ai-draw-io
npm install
cp env.example .env.local # add your provider and API key, see "Models and providers" below
npm run dev
Open http://localhost:6002.
One-click deploy
| Platform | How |
|---|---|
| Tencent EdgeOne Pages |  Deploying there also gives you a daily free quota for DeepSeek models |
| Vercel | .env.local |
| Cloudflare Workers | Cloudflare deploy guide |
| Docker | Docker guide |
| Offline or intranet | Offline deployment |
Models and providers
25 providers are supported: AWS Bedrock (default), OpenAI, Anthropic, Google AI, Google Vertex AI, Azure OpenAI, Ollama, OpenRouter, AIHubMix, DeepSeek, SiliconFlow, SGLang, Vercel AI Gateway, Tencent EdgeOne, ByteDance Doubao, ModelScope, Zhipu GLM, Qwen, Qiniu, Kimi, MiniMax, Novita, Xiaomi MiMo, Atlas Cloud, and an OpenAI-compatible preset for any other URL (vLLM, LM Studio, a gateway). The environment variables and notes for each are in the provider configuration guide.
Which model: the task is to produce long text in a strict format (draw.io XML), so pick a capable model; small models tend to produce broken diagrams.
Several models and the admin panel: list several model IDs in AI_MODEL separated by commas, or configure models from several providers with the AI_MODELS_CONFIG environment variable or an ai-models.json file, and every user can use them without a key of their own. Set ADMIN_PASSWORD and open /admin to manage models, access codes, feature switches, observability and quotas from a web page; see the admin panel guide.
Support
- Questions and ideas: open a GitHub issue or email me[at]jiang.jp
- Common problems: FAQ
- If the project is useful to you, consider sponsoring to help keep the demo site running