drawio-desktop-mcp

drawio-desktop-mcp

Enables AI agents to create and edit diagrams in the draw.io desktop app through the Model Context Protocol.

Category
访问服务器

README

About

drawio-desktop-mcp is a fork of drawio-desktop with a built-in MCP (Model Context Protocol) server that lets AI agents create and edit diagrams directly in the running desktop app.

drawio-desktop is a diagramming desktop app based on Electron that wraps the core draw.io editor.

MCP Server

The app includes an embedded MCP server on http://127.0.0.1:9100/mcp (Streamable HTTP transport) that starts automatically. Multiple AI agents can connect simultaneously and manipulate diagrams in real time.

Quick Start

# Clone (recursive for submodule)
git clone --recursive https://github.com/ArdaUzun/drawio-desktop-mcp.git
cd drawio-desktop-mcp

# Install and run
pnpm install
pnpm start

The MCP server starts automatically on port 9100.

Connect Claude Code

claude mcp add drawio-desktop --transport http http://127.0.0.1:9100/mcp --scope user

Or add a .mcp.json to your project:

{
  "mcpServers": {
    "drawio-desktop": {
      "type": "http",
      "url": "http://127.0.0.1:9100/mcp"
    }
  }
}

Available Tools (20)

Category Tools
Inspect get-diagram-info, get-diagram-screenshot, get-selected-cell, list-paged-model
Shapes add-rectangle, add-cell-of-shape, edit-cell, delete-cell-by-id
Edges add-edge, edit-edge
Shape Library get-shape-categories, get-shapes-in-category, get-shape-by-name, set-cell-shape
Layers list-layers, get-active-layer, set-active-layer, create-layer, move-cell-to-layer
Data set-cell-data

Health Check

curl http://127.0.0.1:9100/health
# {"status":"ok","tools":20,"windows":1,"pending":0}

Flatpak Install

# Build unpacked electron app
pnpm run build:dir
pnpm run build:flatpak-prep

# Build and install Flatpak (requires flatpak-builder + org.electronjs.Electron2.BaseApp//24.08)
flatpak-builder --user --install --force-clean build-flatpak flatpak/com.jgraph.drawio.desktop.yml
flatpak run com.jgraph.drawio.desktop//master

Configuration

Setting CLI Flag Env Var Default
MCP Port --mcp-port <n> DRAWIO_MCP_PORT 9100

Updating from Upstream

git fetch upstream
git merge upstream/dev
git push origin dev

Download built binaries from the releases section.

Can I use this app for free? Yes, under the apache 2.0 license. If you don't change the code and accept it is provided "as-is", you can use it for any purpose.

Security

draw.io Desktop is designed to be completely isolated from the Internet, apart from the update process. This checks github.com at startup for a newer version and downloads it from an AWS S3 bucket owned by Github. All JavaScript files are self-contained, the Content Security Policy forbids running remotely loaded JavaScript.

No diagram data is ever sent externally, nor do we send any analytics about app usage externally. There is a Content Security Policy in place on the web part of the interface to ensure external transmission cannot happen, even by accident.

Security and isolating the app are the primarily objectives of draw.io desktop. If you ask for anything that involves external connections enabled in the app by default, the answer will be no.

Support

Support is provided on a reasonable business constraints basis, but without anything contractually binding. All support is provided via this repo. There is no private ticketing support for non-paying users.

Purchasing draw.io for Confluence or Jira does not entitle you to commercial support for draw.io desktop.

Developing

draw.io is a git submodule of drawio-desktop. To get both you need to clone recursively:

git clone --recursive https://github.com/jgraph/drawio-desktop.git

To run this:

  1. npm install (in the root directory of this repo)
  2. [internal use only] export DRAWIO_ENV=dev if you want to develop/debug in dev mode.
  3. npm start in the root directory of this repo runs the app. For debugging, use npm start --enable-logging.

Note: If a symlink is used to refer to drawio repo (instead of the submodule), then symlink the node_modules directory inside drawio/src/main/webapp also.

To release:

  1. Update the draw.io sub-module and push the change. Add version tag before pushing to origin.
  2. Wait for the builds to complete (https://travis-ci.org/jgraph/drawio-desktop and https://ci.appveyor.com/project/davidjgraph/drawio-desktop)
  3. Go to https://github.com/jgraph/drawio-desktop/releases, edit the preview release.
  4. Download the windows exe and windows portable, sign them using signtool sign /a /tr http://rfc3161timestamp.globalsign.com/advanced /td SHA256 c:/path/to/your/file.exe
  5. Re-upload signed file as draw.io-windows-installer-x.y.z.exe and draw.io-windows-no-installer-x.y.z.exe
  6. Add release notes
  7. Publish release

Note: In Windows release, when using both x64 and is32 as arch, the result is one big file with both archs. This is why we split them.

Local Storage and Session Storage is stored in the AppData folder:

  • macOS: ~/Library/Application Support/draw.io
  • Windows: C:\Users\<USER-NAME>\AppData\Roaming\draw.io\

Not open-contribution

draw.io is closed to contributions (unless a maintainer permits it, which is extremely rare).

The level of complexity of this project means that even simple changes can break a lot of other moving parts. The amount of testing required is far more than it first seems. If we were to receive a PR, we'd have to basically throw it away and write it how we want it to be implemented.

We are grateful for community involvement, bug reports, & feature requests. We do not wish to come off as anything but welcoming, however, we've made the decision to keep this project closed to contributions for the long term viability of the project.

推荐服务器

Baidu Map

Baidu Map

百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。

官方
精选
JavaScript
Playwright MCP Server

Playwright MCP Server

一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。

官方
精选
TypeScript
Magic Component Platform (MCP)

Magic Component Platform (MCP)

一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。

官方
精选
本地
TypeScript
Audiense Insights MCP Server

Audiense Insights MCP Server

通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。

官方
精选
本地
TypeScript
VeyraX

VeyraX

一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。

官方
精选
本地
graphlit-mcp-server

graphlit-mcp-server

模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。

官方
精选
TypeScript
Kagi MCP Server

Kagi MCP Server

一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。

官方
精选
Python
e2b-mcp-server

e2b-mcp-server

使用 MCP 通过 e2b 运行代码。

官方
精选
Neon MCP Server

Neon MCP Server

用于与 Neon 管理 API 和数据库交互的 MCP 服务器

官方
精选
Exa MCP Server

Exa MCP Server

模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。

官方
精选