Docling
Open source document parser that runs locally on macOS, Linux, and Windows, with OCR and structured exports for AI applications.
Docling is an MIT-licensed, open source document parser for developers turning files into structured content for search and AI applications. It runs locally on macOS, Linux, and Windows, including in air-gapped environments. Its PDF processing identifies page layout and reading order, extracts tables, code, and formulas, and classifies images.
It handles scanned files too. OCR reads scanned PDFs and images. Docling also parses DOCX, PPTX, XLSX, HTML, EPUB, Apple Pages, email, and OpenDocument files. Speech recognition supports audio files, while video parsing produces a transcript and representative keyframes. For bar, pie, and line charts, it can produce tables or code with descriptions.
Different file types feed into a shared DoclingDocument structure, which makes their contents easier to use in the same application. Docling can export Markdown, HTML, WebVTT, and lossless JSON, among other formats. It also handles specialized XML schemas for patents, journal articles, and XBRL financial reports, and supports visual language models such as GraniteDocling.
For retrieval and agent applications, Docling integrates with LangChain, LlamaIndex, Crew AI, and Haystack. An MCP server connects it to agents, and the separate docling-serve API server lets teams run it as a service. Developers can also use it through a Python library or command-line interface.
