👁
npm package

Give Claude Code
Eyes

MCP server that lets Claude see your screen. Take screenshots, capture windows, extract text with OCR. Just say "look at my screen."

View on GitHub npm Package
Terminal
# Install globally
$ npm install -g claude-vision
# Add to Claude Code
$ claude mcp add claude-vision -- claude-vision
✓ Added MCP server "claude-vision"
# Now just ask Claude...
"Look at my screen"
"Take a screenshot of VS Code"
"Read the text in this error message"

// CAPABILITIES

What It Can Do

🖥

Multi-Monitor

Capture any connected monitor. List displays and screenshot specific screens by index or name.

🗂

Window Capture

Screenshot specific application windows by title. Grab just VS Code, Chrome, or any app.

▣

Region Select

Capture a specific rectangular region of the screen with precise coordinates.

🔐

OCR Extraction

Extract readable text from screenshots. Perfect for error messages and terminal output.

⏱

Delayed Capture

Set a delay before capture to grab menus, tooltips, and hover states.

🗑

Auto Cleanup

Automatically manage temporary screenshot files to save disk space.

// INSTALLATION

Get Started in 30 Seconds

Step 1: Install

npm install -g claude-vision

Step 2: Add to Claude

claude mcp add claude-vision -- claude-vision

Step 3: Use It

"Look at my screen" just works now

// USE CASES

Built For Developers

Debug Visually

"Look at this error dialog and help me fix it"

UI Review

"Screenshot my app and suggest design improvements"

Documentation

"Capture this flow and write documentation for it"

Code Review

"Look at VS Code and review the highlighted code"

Open Source & Free

MIT licensed. Contributions welcome.

Star on GitHub