Screenshot Analysis MCP Server

Real MCP (Model Context Protocol) server for AI-powered screenshot analysis. Connect directly from Claude Desktop, Cursor, or any MCP client.

Connect from Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "screenshot-analysis": {
      "url": "https://ntriqpro--screenshot-analysis-mcp.apify.actor/mcp"
    }
  }
}

Restart Claude Desktop. The 5 tools will be available immediately.

5 Tools

Tool	Description	Price
analyze_ui	Analyze layout, elements, UX quality	$0.10
compare_screenshots	Visual diff between two screenshots	$0.15
check_accessibility	WCAG accessibility issues	$0.10
extract_text	OCR text extraction	$0.05
describe_page	Detailed page description for documentation	$0.08

Tool Descriptions

1. analyze_ui

Comprehensive UI analysis focusing on layout, elements, and user experience quality.

Input:

• imageUrl (string): URL of the screenshot

Output:

• Layout analysis (grid structure, whitespace, alignment, hierarchy)
• Element inventory (buttons, forms, cards, navigation, etc.)
• UX quality assessment
• Potential issues (clutter, unclear CTAs, navigation problems)
• Strengths of the design

Example Use Case: Review a competitor's website design or audit your own UI for layout issues.

2. compare_screenshots

Detects visual changes between two screenshots.

Input:

• imageUrl1 (string): URL of the first screenshot (before)
• imageUrl2 (string): URL of the second screenshot (after)

Output:

• Before state analysis
• After state analysis
• Comparison note highlighting changes in sections, elements, colors, and layout

Example Use Case: Verify design changes were applied correctly, track UI evolution over versions.

3. check_accessibility

Checks for WCAG accessibility compliance issues in the screenshot.

Input:

• imageUrl (string): URL of the screenshot

Output:

• Color contrast issues
• Readability problems (font size, spacing)
• Touch target sizing issues
• Color dependence issues
• Visual hierarchy clarity
• Spacing and motor accessibility
• Text clarity problems
• Focus indicator visibility

Each issue includes severity level (critical/moderate/minor) and suggested fixes.

⚠️ Important Disclaimer:

• Automated analysis catches ~30-40% of WCAG issues
• Manual testing by accessibility experts is required for compliance
• This tool is informational only, not a compliance certification

Example Use Case: Catch common accessibility issues before formal audit, guide remediation priorities.

4. extract_text

Extracts all visible text from a screenshot using OCR, preserving layout structure.

Input:

• imageUrl (string): URL of the screenshot

Output:

• All visible text organized by reading order (top-left to bottom-right)
• Includes headings, body text, button labels, form labels, navigation items

Example Use Case: Quickly copy text from a screenshot, extract content for translation, document UI copy.

5. describe_page

Generates comprehensive descriptions of page/app functionality useful for documentation and accessibility.

Input:

• imageUrl (string): URL of the screenshot
• context (string, optional): Additional context (e.g., "user trying to reset password")

Output:

• Page title/header
• Main purpose
• Overall layout description
• Navigation structure
• Key content areas
• Interactive elements
• Visual design details
• Call-to-action elements
• Specific data/information displayed
• Context clues (breadcrumbs, status indicators)

Example Use Case: Create documentation for new features, generate alt text for accessibility, share page purpose with non-technical stakeholders.

How It Works

This is a real MCP server using:

• Apify Standby mode (always-on, no cold start)
• Streamable HTTP transport (/mcp endpoint)
• @modelcontextprotocol/sdk for protocol compliance
• Local AI vision models (ntriq vision server)

Unlike regular Apify actors that require API calls, this server speaks native MCP protocol — Claude Desktop connects directly.

Vision API Backend

All analysis is powered by local AI vision models running on ntriq vision server (https://vision.ntriq.co.kr/analyze/image).

Advantages:

• No external API calls for privacy-sensitive screenshots
• Zero data leaves your infrastructure
• Fast processing (local models)
• Consistent analysis quality

Pricing

Pay per event. Only charged when a tool executes successfully. Platform usage is free.

Tool	Cost
analyze_ui	$0.10 per run
compare_screenshots	$0.15 per run
check_accessibility	$0.10 per run
extract_text	$0.05 per run
describe_page	$0.08 per run

Usage Policy

⚠️ Legal Notice:

• Analyze only screenshots you OWN or have explicit permission to analyze
• Do not analyze screenshots containing sensitive personal information, credentials, or copyrighted material
• Automated accessibility analysis is informational only — it catches ~30-40% of WCAG issues
• Manual testing by qualified accessibility experts is required for legal WCAG compliance
• This tool is for analysis and documentation purposes only

Workflow Examples

Example 1: Design Review

1. Take screenshot of your new feature
2. Run analyze_ui to get detailed layout feedback
3. Run check_accessibility to find issues before launch
4. Use describe_page to document the feature

Example 2: A/B Testing

1. Take screenshot of Design A
2. Take screenshot of Design B
3. Run compare_screenshots to see differences
4. Run analyze_ui on both to evaluate UX quality

Example 3: Accessibility Audit

1. Take screenshots of key pages
2. Run check_accessibility on each
3. Document issues found
4. Use describe_page for detailed context
5. Manually verify with accessibility experts

Example 4: Documentation

1. Take screenshot of feature
2. Run extract_text to get UI labels
3. Run describe_page to write detailed description
4. Combine for comprehensive feature docs

Integration with Claude

Claude can use these tools in conversations to:

• Analyze designs you share
• Compare UI changes
• Find accessibility issues
• Extract content
• Generate documentation
• Review competitors' designs

Share a screenshot URL, ask Claude to analyze it, and get instant insights.

Troubleshooting

"Vision server error"

• Check that ntriq vision server is running
• Verify NTRIQ_VISION_URL environment variable if customized
• Ensure image URL is publicly accessible

Empty response

• Image URL may not be publicly accessible
• Large images may timeout (consider resizing)
• Vision server may be temporarily unavailable

Accessibility check seems incomplete

• Remember: automated analysis is ~30-40% accurate
• Always follow up with manual accessibility testing
• Use tools like WAVE, Axe, or NVDA for comprehensive audits

Data Privacy

• Images are sent to ntriq vision server for analysis
• Analysis results are returned immediately
• No image storage or archiving
• Local on-premises deployment available upon request

Support

For issues, feedback, or feature requests, please contact ntriq support.

---

Version: 1.0.0 Categories: MCP Servers, AI, Developer Tools License: Commercial (Apify)