Screeny

This macOS MCP server provides window enumeration and screenshot capture capabilities for specific application windows, enabling AI agents to take visual screenshots of running applications for development and debugging tasks. Built by Rohan Ravindran using Python with FastMCP, pyobjc-framework-Quartz, and Pillow, it offers two core tools: list_windows for discovering available application windows with real window IDs, and take_screenshot for capturing specific windows with optional compression to reduce file sizes for transmission. The implementation features a security-first approach requiring explicit window approval through an interactive setup process, intelligent image compression with multiple quality strategies to stay under MCP size limits, and direct integration with macOS Quartz framework for reliable window detection and screencapture utility for high-quality captures. It handles edge cases like minimized windows, closed applications, and permission requirements, making it valuable for automated testing workflows, debugging visual applications, documentation generation, and building AI assistants that need visual context of running applications without manual screenshot processes.

Source

Repository: https://github.com/rohanrav/screeny