Summary
Whatever is on screen becomes text without a copy, a paste or an upload.
One tool with two arguments, and the left-half default tells you what it was built for: a side-by-side layout where the model reads one half while you work in the other. There is no rectangle selection and no window targeting — halves or the whole screen is the entire choice.
What it is
A single-tool screen capture with OCR: one call takes the shot, reads the text and saves the image.
What you get
- A capture of the left half, the right half or the full screen, selected with region and defaulting to left
- OCR text returned as json, markdown, vertical or horizontal, defaulting to markdown
- The screenshot saved to a dated directory in Downloads
Requirements
No account and no key. The capture is taken from the screen of the machine the server runs on.
Setup effort
One command — npx -y @kazuph/mcp-screenshot