PaddleOCR

This MCP server provides AI assistants with PaddleOCR text extraction and document structure parsing capabilities, built by PaddlePaddle Authors using Python with FastMCP and supporting both local library execution and remote service integration. The implementation offers two core tools: OCR for extracting text from images and PDFs with confidence scores and bounding box coordinates, and PP-StructureV3 for parsing complex documents into structured markdown with embedded images, tables, and formulas. Built with flexible deployment options supporting local PaddleOCR libraries, AI Studio cloud services, and self-hosted servers, plus multiple input formats including file paths, URLs, and Base64 data, it serves document digitization workflows, content management systems requiring text extraction, and research applications needing structured document analysis with precise layout preservation.