Luma

•
116
•

Provides visual understanding capabilities to AI assistants that lack native image processing by integrating multiple vision models.

Luma is a multi-model vision understanding server designed to augment AI assistants that do not inherently support image comprehension. It acts as an MCP (Model Context Protocol) server, enabling AI tools like Claude Desktop, Cline, and Claude Code to perform visual analysis tasks. Luma supports powerful models such as GLM-4.5V for excellent Chinese understanding, DeepSeek-OCR for robust and free OCR capabilities, and Qwen3-VL-Flash for fast, cost-effective processing. It intelligently handles various scenarios including code screenshots, UI analysis, error diagnosis, and general OCR text recognition, making vision capabilities accessible to a wider range of AI applications.

主要功能

01Multi-model support for GLM-4.5V, DeepSeek-OCR (free), and Qwen3-VL-Flash
020 GitHub stars
03Standard MCP protocol for seamless integration with clients like Claude Desktop and Cline
04Intelligent scene recognition for code, UI, and errors
05Unified `analyze_image` tool for all image analysis tasks
06Support for local files, remote image URLs, and Data URIs with built-in retry mechanisms

使用案例

01Analyzing code screenshots for errors and suggestions
02Evaluating UI designs for layout and usability issues
03Performing OCR for text recognition in images to extract information