A drop-in MCP server that pairs long-context reasoning LLMs with vision models in description-only mode, enabling any reasoning model to 'see' images without the vision model giving advice or solution…