Docling is an open-source document parsing framework developed by IBM Research Zurich and hosted under the LF AI & Data Foundation. It is designed to facilitate the ingestion of diverse file formats—including PDF, DOCX, XLSX, HTML, EPUB, and various image formats—into a unified, expressive DoclingDocument representation. The tool provides advanced PDF comprehension, capable of extracting page layouts, reading orders, complex table structures, formulas, and code. Beyond standard text, it features specialized capabilities such as chemical molecular structure recognition, chart understanding (converting bar, pie, and line plots into tables or code), and audio/video processing via Automatic Speech Recognition (ASR) for generating transcripts and keyframes. For visual intelligence, it supports integration with Visual Language Models (VLMs) like GraniteDocling. Developers can export processed data into several formats, including Markdown, HTML, and lossless JSON, or utilize application-specific XML schemas like USPTO patents and XBRL financial reports. Docling offers flexible deployment through a CLI, an API server (docling-serve), or as an MCP server, and features native integrations with AI orchestration frameworks such as LangChain, LlamaIndex, CrewAI, and Haystack. The library requires Python 3.10 or higher and supports x86_64 and arm64 architectures across macOS, Linux, and Windows.