This MCP server provides fast, high-quality webpage screenshots tailored for AI vision workflows. It uses a Puppeteer headless browser to capture full-page content, applying automatic resolution limiting to 1072x1072 pixels (approximately 1.15 megapixels) to optimize processing by Claude Vision API and other AI models. Full-page captures are tiled into 1072x1072 tiles, enabling efficient feeding into AI pipelines. The server supports screencasts by capturing a series of screenshots over time with configurable intervals, and can export animated WebP files for compact, high-quality motion. It ensures content is always fresh with no caching, and offers configurable viewports for responsive testing. Wait strategies for dynamic content (load, domcontentloaded, networkidle0, networkidle2) and optional delays are supported. By default, it captures the full page, but outputs can be saved to a directory (returning file paths) or used in-memory. Optional JavaScript injection allows custom code to run before screencast capture. Additional features include minimal dependencies, MCP integration, automatic browser cleanup after 60 seconds of inactivity, and closing pages after each screenshot to prevent memory leaks. This makes it ideal for integrating reliable, AI-ready web captures into MCP workflows.