MCP server that analyzes images with Google's Gemini vision models, allowing agents to describe or ask questions about images without bloating context.