# Shadow Vision MCP server

> Enables text-only LLMs to perform OCR, visual analysis, UI inspection, and image reconstruction using pluggable vision backends.

- Repository: https://github.com/WardLu/shadow-vision
- Page: https://hostedmcp.org/servers/wardlu/shadow-vision
- Transport: —
- Stars: 1
- License: MIT

## Environment variables

| Name | Required | Secret | Description |
|---|---|---|---|
| `VISION_BACKEND` | no | no |  |

## Tools (3)

- `task`
- `image_path`
- `image_url`

## FAQ

### How can I keep my images private?

You can use the local Ollama backend to ensure images and inference remain entirely on your machine, preventing external data transfer.

### Can I use this with my existing text-only LLM?

Yes, it is designed to give text-only LLMs visual capabilities like OCR and UI inspection without needing to switch the host model.
