Screen View

You use the Screen View MCP Server to capture and analyze screenshots with AI insights, enabling you to understand UI layouts, debug visual issues, and document screen content. It supports both local, stdio-based operation and SSE transport for remote or more interactive setups.

How to use

Start the Screen View MCP Server locally or via a remote setup, then connect your MCP client to invoke the captureAndAnalyzeScreen tool. You can trigger a full-screen capture, request AI-driven analysis of the current screen, and optionally save the screenshot for later reference. Use prompts to guide the analysis and extract layout details, element positions, and visual structure. If you need to run the server on a different machine, you can expose an SSE endpoint and connect your client to that endpoint.

Typical workflow:

• Install the MCP package globally on your system.
• Start the local stdio server (or connect to a remote SSE server).
• Invoke captureAndAnalyzeScreen from your MCP client with an optional prompt and model preference.
• Review the AI-generated description of the current screen and any saved screenshots.

How to install

Install the Screen View MCP Server globally using npm to ensure you have the latest stable release.

# Install the latest version
npm install -g screen-view-mcp

# To ensure you get the exact latest version and avoid caching issues
npm install -g screen-view-mcp@2.0.15  # Replace with latest version number

Configuration and start options

Choose a transport method that matches your client. For Claude Desktop and other stdio-capable clients, use stdio by default. If your client supports SSE, you can enable it for remote access.

{
  "mcpServers": {
    "screen-view-mcp": {
      "command": "npx",
      "args": [
        "screen-view-mcp@2.0.15"
      ],
      "transport": "stdio",
      "env": {
        "ANTHROPIC_API_KEY": "your-anthropic-api-key"
      }
    }
  }
}

Advanced transport: SSE mode

If you need SSE transport for local or remote connections, you can enable it and expose a port for the client to connect.

{
  "mcpServers": {
    "screen-view-mcp": {
      "command": "npx",
      "args": [
        "screen-view-mcp@2.0.15",
        "--sse",
        "--port", "8080",
        "--host", "localhost"
      ],
      "env": {
        "ANTHROPIC_API_KEY": "your-anthropic-api-key"
      }
    }
  }
}

Connecting to a remote SSE server

You can connect to a remote SSE server by providing the URL of the SSE endpoint and using the SSE transport.

{
  "mcpServers": {
    "screen-view-mcp": {
      "url": "http://your-server-ip:8080/sse",
      "transport": "sse",
      "env": {
        "ANTHROPIC_API_KEY": "your-anthropic-api-key"
      }
    }
  }
}

Tools and usage

The SCP server provides a tool to capture and analyze the current screen. You can supply a prompt to guide the analysis and select the Claude model for analysis.

Troubleshooting

Common issues include lack of screen capture permissions, invalid API keys, missing MCP package, or transport mismatches. Ensure you have screen capture permissions enabled for your AI client, verify the Anthropic API key, install the MCP package globally, and use the correct transport mode for your client.

Development notes

To run locally for testing, install dependencies, build if needed, and start with the appropriate command. The following commands illustrate testing with stdio and SSE transports.

# Build and test locally (example commands)
npm install
npm run build

# Test with stdio transport (default)
node dist/screen-capture-mcp.js --api-key=your-anthropic-api-key

# Test with SSE transport
node dist/screen-capture-mcp.js --sse --port 8080 --host localhost --api-key=your-anthropic-api-key

Available tools

captureAndAnalyzeScreen

Captures the current screen content and analyzes it with AI, returning a description of the layout and elements on the screen.

helloWorld

Simple test tool that echoes a message to verify MCP connectivity and basic behavior.