A screenshot from a modern monitor is far too large to hand to a model directly. This one steps the image down progressively — 1920px, then 1280px, then 800px — until it fits under 1MB, adjusting quality along the way, which is why a full-desktop capture returns rather than failing. If a capture still will not come back, lowering the quality parameter is the documented first move. Everything is processed locally with no network transmission.
A capture bridge for developers working in WSL against a Windows desktop. It drives PowerShell to grab monitors, windows or the clipboard, then processes the image so it can actually be returned to a model.
- `take_screenshot` captures all monitors, the primary monitor, or a specific monitor by number
- The same tool targets a window by title with partial matching, or by process name, with `windowIndex` picking between several matches
- `returnDirect` chooses between handing the image straight to the assistant and saving it, with `filename`, `folder` and a `quality` setting that defaults to 80
- `read_clipboard` auto-detects whether the clipboard holds text or an image, with `format` to force one
- Multi-monitor setups with different scaling factors are handled correctly
- Paths work in both WSL and Windows form
Windows 10 or 11 with WSL2, Node.js 18.0.0 or higher, and PowerShell, which ships with Windows. Clone and install; the client launches it through `wsl --cd` pointed at the install directory. Published as wslsnapit-mcp, version 2.2.0.
One command — wsl --cd /mnt/h/dev/mcp/wslsnapit-mcp node index.js