Enables analysis of local images through Kimi (Moonshot AI) vision models via the MCP protocol, supporting features like OCR and long context understanding.
Enables automated grading of student homework images using Qwen3-VL multimodal model, supporting multiple subjects and question types with detailed feedback and batch processing capabilities.
Enables AI assistants to recognize and extract information from images via GLM-4V, supporting automatic screenshot recognition and MCP-based local image file reading for non-vision models like DeepSeek.
Enables AI assistants and coding models without native vision to analyze images—screenshots, error messages, UIs, documents and OCR text—through a single unified image_understand tool. Supports multiple vision LLM providers (GLM-4.6V, DeepSeek-OCR, Qwen3-VL, Doubao, Hunyuan, or any OpenAI-compatible endpoint), with automatic image preprocessing, cropping and retries, over stdio or shared HTTP/Docker deployment.