Brings MinerU's document parsing engine to Claude Desktop and other MCP clients, letting AI assistants convert PDFs, Office files, images, and web pages into clean Markdown on demand. Exposes both flash_extract for quick, token-free parsing under 10MB and extract for precision work with OCR, LaTeX formulas, and table reconstruction. Built on the MinerU Open API, which handles 109 languages and preserves complex layouts including multi-column text and cross-page tables. Useful when you want Claude to read documents during a conversation without preprocessing files yourself. The underlying engine powers production RAG pipelines and was designed for LLM training workflows, so output quality is tuned for downstream language model consumption.
claude mcp add --transport stdio io.github.opendatalab-mineru-open-mcp uvx mineru-open-mcp