SCCPT - Python Screen Capture with MCP Server Support

- shared visual context between users and AI agents -

A lightweight, efficient screen capture library with automatic error detection. Features full MCP (Model Context Protocol) server integration for seamless AI assistant workflows. Optimized for WSL2 to capture Windows host screens.

Key Features

🤖 MCP Server Integration

• AI Assistant Ready - Built-in MCP server for Claude Code and other AI assistants
• Direct Screenshot Control - AI can capture, monitor, and analyze screenshots programmatically
• Automated Workflows - Perfect for debugging, documentation, and monitoring tasks
• Real-time Interaction - AI assistants can respond to visual changes instantly

📸 Core Screen Capture

• 🖼️ Windows Host Capture from WSL - Seamlessly capture Windows desktop from WSL with DPI awareness
• 🖥️ Multi-Monitor Support - Capture single monitor or all monitors combined
• 🚨 Automatic Error Detection - Automatically categorizes screenshots when called in exception handlers
• 🗜️ JPEG Compression - Reduce file sizes with configurable quality (~250KB JPEG vs ~2MB PNG)

🎬 Advanced Features

• ⏱️ Continuous Monitoring - Automatic captures at configurable intervals (WSL: ~1.5s minimum due to PowerShell overhead)
• 🎥 GIF Summaries - Create animated GIFs from monitoring sessions for easy sharing
• 📝 Smart Filenames - Timestamps + normalized messages + category indicators
• 🔄 Thread-Safe - Safe for concurrent operations
• 🎯 Minimal Logging - Clean, single-line output per screenshot

Installation

pip install sccpt
# pip install sccpt[full] # For full features including mss and Pillow support:

Quick Start

🐍 Python API - Beautiful Simplicity (Just 4 Functions)

import sccpt

sccpt.cpt()                          # Capture single screenshot
sccpt.start()                        # Start monitoring  
sccpt.stop()                         # Stop monitoring
sccpt.create_gif_from_latest_session()  # Create GIF summary

🤖 MCP Server - AI Assistant Integration

# Start MCP server for Claude Code integration
python -m mcp_server_sccpt

For complete setup instructions and available tools, see MCP Server Integration section below.

Use Cases & Examples

📸 Essential Use Cases

• Debug visually - Capture before/after states during development
• Monitor processes - Continuous screenshots during long operations
• Document workflows - Step-by-step visual documentation
• Error analysis - Automatic error categorization and screenshots
• AI-powered automation - Let AI assistants capture and analyze screens via MCP integration

import sccpt

def process_data(df):
    sccpt.cpt("before transformation")
    df = df.transform(complex_operation)
    sccpt.cpt("after transformation")
    return df

import sccpt

try:
    selenium_driver.click(button)
    api_response = fetch_data()
except Exception as e:
    sccpt.cpt()  # Auto-adds -stderr suffix
    raise

import sccpt

sccpt.start()  # Start taking screenshots every second
train_model()  # Your long operation
sccpt.stop()

import sccpt

sccpt.start()
# ... your process ...  
sccpt.stop()
sccpt.create_gif_from_latest_session()
# 📹 GIF created: ~/.cache/sccpt/20250823_104523_summary.gif

Platform Support

sccpt.start(
    output_dir="~/screenshots",  # Where to save
    interval=2.0,                # Seconds between captures
    quality=85,                  # JPEG quality (1-100)
    verbose=False                # Silent mode
)

~/.cache/sccpt/
├── 20250823_104523-message-stdout.jpg    # Normal capture
├── 20250823_104525-error-stderr.jpg      # Error capture  
└── 20250823_104530_0001_*.jpg            # Monitoring mode

Cache automatically managed (1GB default limit, oldest files removed)

🤖 MCP Server Integration

Transform your AI workflows - SCCPT's MCP server enables Claude and other AI assistants to capture, monitor, and analyze screenshots directly as part of their reasoning process.

AI-Powered Capabilities

• Seamless Integration - AI assistants can capture screenshots on demand
• Automated Debugging - AI captures error screens and analyzes issues
• Documentation Generation - AI creates visual step-by-step guides
• Intelligent Monitoring - AI watches for visual changes and responds
• Error Analysis - AI automatically categorizes and analyzes screenshots

// Add to your Claude Code settings
{
  "mcpServers": {
    "sccpt": {
      "command": "python", 
      "args": ["/path/to/sccpt/mcp_server_sccpt.py"]
    }
  }
}

🎬 Live Demos & Screenshots

📸 Single Screenshot Capture

Example of SCCPT's single screenshot capture functionality with custom message

🔄 Real Monitoring Session

Real GIF created from an actual monitoring session showing SCCPT's continuous capture capabilities

Real Session Details:

• Duration: 30 seconds monitoring interval
• Screenshots: 11 captures automatically taken
• File Size: 2.8MB optimized GIF
• Quality: Automatic JPEG compression for efficient storage
• Features Demonstrated:
  • Automatic timestamp-based naming
  • Session-based organization
  • GIF creation from monitoring sessions

🎥 Professional Feature Demonstrations

📋 Complete Workflow Documentation

Step-by-step workflow documentation showing how SCCPT captures each stage of a process (230KB, 7 frames)

🖥️ Continuous Monitoring Visualization

Professional demonstration of SCCPT's continuous monitoring with real-time progress tracking (429KB, 12 frames)

🚨 Automatic Error Detection

Shows how SCCPT automatically categorizes screenshots as stdout (normal) vs stderr (error) based on execution context (322KB, 5 frames)

License

MIT License

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Contact

Yusuke.Watanabe@scitex.ai