design-canvas-mcp
Enables AI agents to manage routes on the Design Canvas infinite canvas app, including adding, removing, listing routes, and capturing screenshots of the canvas.
README
design-canvas-mcp
MCP server for Design Canvas — control the infinite canvas from AI coding agents.
Design Canvas is a native Mac app that shows all your web app's routes on an infinite canvas with live browser views. This MCP server lets AI agents (Claude Code, Cursor) add routes, capture screenshots, and manage what's displayed.
Prerequisites
- Design Canvas Mac app must be running on your machine
- The app must be connected to a running dev server (e.g.,
localhost:3000) - Node.js 20+
Installation
Claude Code
claude mcp add design-canvas -s user -- npx -y design-canvas-mcp
Cursor
Add to .cursor/mcp.json:
{
"mcpServers": {
"design-canvas": {
"command": "npx",
"args": ["-y", "design-canvas-mcp"]
}
}
}
Global install (alternative)
npm install -g design-canvas-mcp
Then configure with command: "design-canvas-mcp" instead of the npx variant.
Tools
| Tool | Description | Key Parameters |
|---|---|---|
list_routes |
List all routes on the canvas with IDs, paths, dimensions, positions | — |
add_route |
Add a route to the canvas from the connected dev server | path (required), width (1280/768/375) |
remove_route |
Remove a route frame by ID or path | id (preferred) or path |
capture_canvas |
Capture screenshots of routes and return as images | ids (optional), limit (default: 3) |
Example prompts
- "Show me all routes on the canvas" →
list_routes - "Add /dashboard at desktop and mobile" →
add_routeat 1280 + 375 - "Capture the canvas so you can see my app" →
capture_canvas - "Remove the mobile version of /dashboard" →
list_routesthenremove_routeby ID
How it works
The MCP server communicates with the Design Canvas Mac app via a local HTTP API on localhost:7420. The server is an adapter layer — it translates between the HTTP API and MCP's content block format.
Claude Code / Cursor
→ spawns design-canvas-mcp (stdio)
→ HTTP requests to localhost:7420
→ Design Canvas Mac app responds
Screenshots are captured by the Mac app's WebKit rendering engine, resized to 1200px width, and returned as base64 JPEG images.
Troubleshooting
"Design Canvas is not running" The Mac app isn't open or isn't listening on port 7420. Open Design Canvas and make sure it shows the canvas view.
Routes show but pages are blank
The dev server isn't running. Start your dev server (e.g., npm run dev) and make sure Design Canvas is connected to it.
Screenshots are from WebKit, not Chrome Design Canvas uses WebKit (Safari's engine) for rendering. For pixel-perfect Chromium verification, use Playwright. The MCP screenshots are for quick layout checks, not cross-browser QA.
License
MIT
推荐服务器
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。
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。
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。
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。
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。
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。
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。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
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。
mcp-server-qdrant
这个仓库展示了如何为向量搜索引擎 Qdrant 创建一个 MCP (Managed Control Plane) 服务器的示例。