Product / UX / Agent tools & protocols

Docling

SDK and CLI that parses PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, audio and video into unified structured output - Markdown, lossless JSON, HTML, WebVTT and DocTags - for generative-AI and RAG pipelines. Unlike URL-to-Markdown services it does deep local document understanding: layout and reading-order analysis, table-structure recognition, formula and code extraction, OCR, and vision and speech model integration. It also reads application-specific XML schemas including USPTO patents, JATS articles and XBRL filings, and ships its own MCP server.

Originated at IBM Research, inducted into the Linux Foundation AI & Data Foundation (Docling Project) on 2025-04-29 under community governance. Verified 2026-08-13 via GitHub, the repository LICENSE and the pypistats API.

Openness

5 high confidence

5.0

license: MIT(OSI)
source: public(github.com/docling-project/docling)
self-host: primary
core-gated: ungated

MIT, local-first document parsing, fully self-hostable with integrated open-weight models.

https://raw.githubusercontent.com/docling-project/docling/main/LICENSE recorded 2026-08-13
Plain MIT License text: "Copyright (c) 2024 International Business Machines".
https://github.com/docling-project/docling recorded 2026-08-13
Repo page - MIT badge, 64,719 stargazers. "Local execution capabilities for sensitive data and air-gapped environments"; open-weight VLMs such as GraniteDocling driven from the CLI; docling-serve as the self-run API server. No paid tier, license key or enterprise directory.

Adoption

5 medium confidence

5.0

12,493,111 downloads of docling in the trailing 30 days, which bands at >10M, level 5 on the software adoption scale. The count is for the first-party package alone; docling-core adds more again on top of it.

https://pypistats.org/api/packages/docling/recent recorded 2026-08-12
12,493,111 downloads in the trailing 30 days for docling

Capability

5 high confidence

5.0

Deepest local document understanding among RAG ingestion tools; web-URL parsers lack rich binary-doc depth.

https://github.com/docling-project/docling recorded 2026-08-13
Feature list - PDF, DOCX, PPTX, XLSX, HTML, EPUB, WAV, MP3, WebVTT, Box Notes, EML/MSG, images, LaTeX, DocLang and plain text input; "Advanced PDF understanding incl. page layout, reading order, table structure, code, formulas, image classification"; export to Markdown, HTML, WebVTT, DocLang, DocTags and lossless JSON; "Extensive OCR support for scanned PDFs and images"; VLM support; "Audio support with Automatic Speech Recognition (ASR) models"; "Connect to any agent using the MCP server"; video parsing under What's new.

Verified 2026-08-12