Connects Claude directly to your screen through real-time capture and vision analysis. Exposes screen grab functionality that feeds visual data to Claude's vision capabilities, letting it see windows, read text via OCR, and analyze what's currently displayed. Requires an Anthropic API key since it calls Claude's vision models to process the screenshots. Reach for this when you want Claude to debug UI issues by actually looking at your screen, extract text from images or PDFs you have open, or understand visual context without manually describing or uploading screenshots. Part of a larger NIRA toolkit but works standalone as a PyPI package with stdio transport.
claude mcp add --transport stdio pachote-nira-screen-vision uvx nira-screen-vision