Draw.io MCP Server

Draw.io MCP Server

Enables AI agents to programmatically create, modify, and analyze Draw.io diagrams through the Model Context Protocol. Supports generating architectural diagrams, flowcharts, and visualizations with bidirectional communication between AI systems and Draw.io.

Category
访问服务器

README

Draw.io MCP server

Let's do some Vibe Diagramming with the most wide-spread diagramming tool called Draw.io (Diagrams.net).

Discord channel Build project

Introduction

The Draw.io MCP server is a Model Context Protocol (MCP) implementation that brings powerful diagramming capabilities to AI agentic systems. This integration enables:

  • Seamless Draw.io Integration: Connect your MCP-powered applications with Draw.io's rich diagramming functionality
  • Programmatic Diagram Control: Create, modify, and manage diagram content through MCP commands
  • Intelligent Diagram Analysis: Retrieve detailed information about diagrams and their components for processing by AI agents
  • Agentic System Development: Build sophisticated AI workflows that incorporate visual modeling and diagram automation

As an MCP-compliant tool, it follows the standard protocol for tool integration, making it compatible with any MCP client. This implementation is particularly valuable for creating AI systems that need to:

  • Generate architectural diagrams
  • Visualize complex relationships
  • Annotate technical documentation
  • Create flowcharts and process maps programmatically

The tool supports bidirectional communication, allowing both control of Draw.io instances and extraction of diagram information for further processing by AI agents in your MCP ecosystem.

Requirements

To use the Draw.io MCP server, you'll need:

Core Components

  • Node.js (v20 or higher) - Runtime environment for the MCP server
  • Draw.io MCP Browser Extension - Enables communication between Draw.io and the MCP server

MCP Ecosystem

  • MCP Client (e.g., MCP Inspector) - For testing and debugging the integration
  • LLM with Tools Support - Any language model capable of handling MCP tool calls (e.g., GPT-4, Claude 3, etc.)

Optional for Development

  • pnpm - Preferred package manager
  • Chrome DevTools - For debugging when using --inspect flag

Note: The Draw.io desktop app or web version must be accessible to the system where the MCP server runs.

Configuration

WebSocket Port

The server listens on port 3333 by default for WebSocket connections from the browser extension. You can customize this port using the --extension-port or -p flag.

Default behavior (port 3333):

{
  "mcpServers": {
    "drawio": {
      "command": "npx",
      "args": ["-y", "drawio-mcp-server"]
    }
  }
}

Custom port (e.g., port 8080):

{
  "mcpServers": {
    "drawio": {
      "command": "npx",
      "args": ["-y", "drawio-mcp-server", "--extension-port", "8080"]
    }
  }
}

Note: When using a custom port, ensure the browser extension is configured to connect to the same port.

HTTP Transport Port

The server can expose a streamable HTTP MCP transport on port 3000. Change this using the --http-port flag:

{
  "mcpServers": {
    "drawio": {
      "command": "npx",
      "args": ["-y", "drawio-mcp-server", "--transport", "http", "--http-port", "4000"]
    }
  }
}

Transport Selection

By default only the stdio transport starts. Limit or combine transports with the --transport flag:

  • --transport stdio – start only stdio (CLI-friendly)
  • --transport http – start only the HTTP transport (for remote clients)
  • --transport stdio,http – start both transports

Running the streamable HTTP transport

Use the streamable HTTP transport when you need to reach the MCP server over the network (for example from a remote agent runtime). The Draw.io browser extension is still required, and you must opt in to the HTTP transport.

  1. Start the server with HTTP enabled (optionally alongside stdio):
npx -y drawio-mcp-server --transport http --http-port 3000
# or both: npx -y drawio-mcp-server --transport stdio,http --http-port 4000
  1. Verify the health endpoint:
curl http://localhost:3000/health
# { "status": "ok" }
  1. Point your MCP client to the /mcp endpoint (http://localhost:3000/mcp by default). CORS is enabled for all origins so you can call it from a browser-based client as well.

Installation

Connecting with Claude Desktop

  1. Install Claude Desktop

  2. Open or create the configuration file:

    • macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
    • Windows: %APPDATA%\Claude\claude_desktop_config.json
  3. Update it to include this server:

<details> <summary>Using <code>npm</code></summary>

{
   "mcpServers":{
      "drawio":{
         "command":"npx",
         "args":[
            "-y",
            "drawio-mcp-server"
         ]
      }
   }
}

</details>

<details> <summary>Using <code>pnpm</code></summary>

{
   "mcpServers":{
      "drawio":{
         "command":"pnpm",
         "args":[
            "dlx",
            "drawio-mcp-server"
         ]
      }
   }
}

</details>

To use a custom extension port (e.g., 8080), add "--extension-port", "8080" to the args array:

{
  "mcpServers": {
    "drawio": {
      "command": "npx",
      "args": ["-y", "drawio-mcp-server", "--extension-port", "8080"]
    }
  }
}
  1. Restart Claude Desktop

Connecting with oterm

This is an alternative MCP client in case you like terminal and you plan to connect to your own Ollama instance.

The configuration is usually in: ~/.local/share/oterm/config.json

<details> <summary>Using <code>npm</code></summary>

{
	"mcpServers": {
		"drawio": {
			"command": "npx",
			"args": [
			  "-y",
        "drawio-mcp-server"
			]
		}
	}
}

</details>

<details> <summary>Using <code>pnpm</code></summary>

{
	"mcpServers": {
		"drawio": {
			"command": "pnpm",
			"args": [
			  "dlx",
        "drawio-mcp-server"
			]
		}
	}
}

</details>

To use a custom extension port (e.g., 8080), add "--extension-port", "8080" to the args array:

{
  "mcpServers": {
    "drawio": {
      "command": "npx",
      "args": ["-y", "drawio-mcp-server", "--extension-port", "8080"]
    }
  }
}

Connect with Zed

  1. Open the Zed Preview application.
  2. Click the Assistant (✨) icon in the bottom right corner.
  3. Click Settings in the top right panel of the Assistant.
  4. In the Context Servers section, click + Add Context Server.
  5. Configure with the following:

<details> <summary>Using <code>npm</code></summary>

{
  /// The name of your MCP server
  "drawio": {
    "command": {
      /// The path to the executable
      "path": "npx",
      /// The arguments to pass to the executable
      "args": ["-y","drawio-mcp-server"],
      /// The environment variables to set for the executable
      "env": {}
    }
  }
}

</details>

<details> <summary>Using <code>pnpm</code></summary>

{
  /// The name of your MCP server
  "drawio": {
    "command": {
      /// The path to the executable
      "path": "pnpm",
      /// The arguments to pass to the executable
      "args": ["dlx","drawio-mcp-server"],
      /// The environment variables to set for the executable
      "env": {}
    }
  }
}

</details>

To use a custom extension port (e.g., 8080), add "--extension-port", "8080" to the args array:

{
  /// The name of your MCP server
  "drawio": {
    "command": {
      /// The path to the executable
      "path": "npx",
      /// The arguments to pass to the executable
      "args": ["-y","drawio-mcp-server","--extension-port","8080"],
      /// The environment variables to set for the executable
      "env": {}
    }
  }
}

Connecting with Codex

Edit the configuration usually located in: ~/.codex/config.toml

<details> <summary>Using <code>npm</code></summary>

[mcp_servers.drawio]
command = "npx"
args = [
    "-y",
    "drawio-mcp-server"
]

</details>

<details> <summary>Using <code>pnpm</code></summary>

[mcp_servers.drawio]
command = "pnpm"
args = [
    "dlx",
    "drawio-mcp-server"
]

</details>

To use a custom extension port (e.g., 8080), add "--extension-port", "8080" to the args array:

<details> <summary>Using <code>npm</code></summary>

[mcp_servers.drawio]
command = "npx"
args = [
    "-y",
    "drawio-mcp-server",
    "--extension-port",
    "8080"
]

</details>

To connect to a locally running MCP with Streamable HTTP transport:

[mcp_servers.drawio]
url = "http://localhost:3000/mcp"

Browser Extension Setup

In order to control the Draw.io diagram, you need to install dedicated Browser Extension.

  1. Open Draw.io in your browser
  2. Install the Draw.io MCP Browser Extension from a web store or use other means <p> <a href="https://chrome.google.com/webstore/detail/drawio-mcp-extension/okdbbjbbccdhhfaefmcmekalmmdjjide"> <picture> <source srcset="https://i.imgur.com/XBIE9pk.png" media="(prefers-color-scheme: dark)" /> <img height="58" src="https://i.imgur.com/oGxig2F.png" alt="Chrome Web Store" /></picture

</a> <a href="https://addons.mozilla.org/en-US/firefox/addon/drawio-mcp-extension/"> <picture> <source srcset="https://i.imgur.com/ZluoP7T.png" media="(prefers-color-scheme: dark)" /> <img height="58" src="https://i.imgur.com/4PobQqE.png" alt="Firefox add-ons" /></picture </a> </p>

  1. Ensure it is connected, the Extension icon should indicate green signal overlay <img alt="Extension connected" src="https://raw.githubusercontent.com/lgazo/drawio-mcp-extension/refs/heads/main/public/icon/logo_connected_32.png" />

Important: If you configured the MCP server to use a custom port (not 3333), you must configure the browser extension to use the same port. See the extension documentation for port configuration instructions.

Sponsoring

If you enjoy the project or find it useful, consider supporting its continued development.

lightning invoice:

lightning invoice

lnbc1p5f8wvnpp5kk0qt60waplesw3sjxu7tcqwmdp6ysq570dc4ln52krd3u5nzq6sdp82pshjgr5dusyymrfde4jq4mpd3kx2apq24ek2uscqzpuxqr8pqsp5gvr72xcs883qt4hea6v3u7803stcwfnk5c9w0ykqr9a40qqwnpys9qxpqysgqfzlhm0cz5vqy7wqt7rwpmkacukrk59k89ltd5n642wzru2jn88tyd78gr4y3j6u64k2u4sd4qgavlsnccl986velrg3x0pe95sx7p4sqtatttp

lightning address:

ladislav@blink.sv

<div align="center"> <a href="https://liberapay.com/ladislav/donate"><img alt="Donate using Liberapay" src="https://liberapay.com/assets/widgets/donate.svg"></a> </div>

Features

The Draw.io MCP server provides the following tools for programmatic diagram interaction:

Diagram Inspection Tools

  • get-selected-cell Retrieves the currently selected cell in Draw.io with all its attributes Returns: JSON object containing cell properties (ID, geometry, style, value, etc.)

  • get-shape-categories Retrieves available shape categories from the diagram's library Returns: Array of category objects with their IDs and names

  • get-shapes-in-category Retrieves all shapes in a specified category from the diagram's library Parameters:

    • category_id: Identifier of the category to retrieve shapes from Returns: Array of shape objects with their properties and styles
  • get-shape-by-name Retrieves a specific shape by its name from all available shapes Parameters:

    • shape_name: Name of the shape to retrieve Returns: Shape object including its category and style information
  • list-paged-model Retrieves a paginated view of all cells (vertices and edges) in the current Draw.io diagram. This tool provides access to the complete model data with essential fields only, sanitized to remove circular dependencies and excessive data. It allows to filter based on multiple criteria and attribute boolean logic. Useful for programmatic inspection of diagram structure without overwhelming response sizes.

Diagram Modification Tools

  • add-rectangle Creates a new rectangle shape on the active Draw.io page with customizable properties:

    • Position (x, y coordinates)
    • Dimensions (width, height)
    • Text content
    • Visual style (fill color, stroke, etc. using Draw.io style syntax)
  • add-edge Creates a connection between two cells (vertexes) Parameters:

    • source_id: ID of the source cell
    • target_id: ID of the target cell
    • text: Optional text label for the edge
    • style: Optional style properties for the edge
  • delete-cell-by-id Removes a specified cell from the diagram Parameters:

    • cell_id: ID of the cell to delete
  • add-cell-of-shape Adds a new cell of a specific shape type from the diagram's library Parameters:

    • shape_name: Name of the shape to create
    • x, y: Position coordinates (optional)
    • width, height: Dimensions (optional)
    • text: Optional text content
    • style: Optional additional style properties
  • set-cell-shape Applies a library shape's style to an existing cell Parameters:

    • cell_id: ID of the cell whose appearance should change
    • shape_name: Name of the library shape whose style should be applied
  • set-cell-data Stores or updates a custom attribute on a cell Parameters:

    • cell_id: ID of the cell to update
    • key: Attribute name to set
    • value: Attribute value (stored as a string internally)
  • edit-cell Updates an existing vertex/shape cell in place by ID Parameters:

    • cell_id: ID of the cell whose properties should change (required)
    • text, x, y, width, height, style: Optional fields to update on the cell; omitted properties stay as-is
  • edit-edge Updates an existing edge connection between cells by ID Parameters:

    • cell_id: ID of the edge cell to update (required)
    • text: Optional edge label text
    • source_id, target_id: Optional IDs of new source/target cells
    • style: Optional replacement style string

Related Resources

Troubleshooting

Prompt examples

Contributing

Architecture

Development

Star History

<a href="https://star-history.com/#lgazo/drawio-mcp-server&Date"> <picture> <source media="(prefers-color-scheme: dark)" srcset="https://api.star-history.com/svg?repos=lgazo/drawio-mcp-server&type=Date&theme=dark" /> <source media="(prefers-color-scheme: light)" srcset="https://api.star-history.com/svg?repos=lgazo/drawio-mcp-server&type=Date" /> <img alt="Star History Chart" src="https://api.star-history.com/svg?repos=lgazo/drawio-mcp-server&type=Date" /> </picture> </a>

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