PDF & Documents
SaaS, open source and skills for pdf & documents work — with the limits of each written down.Before you pick
What decides it
Reading order across columns, and whether table cells survive as cells, is decided in the parser before any model sees the page.
What to avoid
Flattened tables come back fluent, correctly cited and numerically wrong; the page reference confirms where it looked, not what it read.
What is changing
Parsing is shifting from OCR text to reading the page as an image, and output from prose answers toward typed fields carrying per-field confidence.
SaaS
Hosted products you sign up for.Leading
Established picks, at the top of today's aggregate.
- 1ChatPDFFast, no-friction chat with a single PDF for quick one-off questions and summaries.
- 2Adobe Acrobat AI AssistantEditing and summarizing existing PDFs with page-level citations plus enterprise security and redaction.
- 3Google NotebookLMBest free option; retrieves source text before answering, so it resists hallucinated citations.
- 4AskYourPDFFree multi-document chat when you need to query across several PDFs at once.
- 5Humata AIResearch-oriented Q&A over long technical and academic PDFs with cited answers.
Emerging
Newer challengers, surfaced by two or more sources.
- 1PDF Vertex
- 2FileVerbsREST API for developers to convert PDF, Word, Excel and image files, with OCR and handling of password-protected documents.
- 3terrific.tools DesktopOffline desktop converter covering 1,000+ format pairs across audio, video, images and documents, processing files locally after a one-time purchase.
- 4PDF Reader ProDesktop and mobile PDF editor for Mac, Windows, iOS and Android, adding forms, OCR and ChatGPT-backed summarise and translate, sold as a one-time purchase.
Open source
Repositories you run yourself.Leading
- 1Stirling-PDFThe de-facto open-source PDF toolkit: 50+ local operations — merge, split, OCR, convert, sign, redact, compress — fully self-hosted. The recognized OSS PDF leader (~88k stars).
- 2MarkItDown (Microsoft)Microsoft's document-to-Markdown converter (PDF, Office, images, audio) built to feed clean text to LLMs; the highest-starred tool in this space (~169k stars).
- 3MinerU (OpenDataLab)High-accuracy PDF-to-Markdown/JSON extraction; the standout for complex academic papers and CJK layouts (~76k stars).
- 4Docling (IBM / DS4SD)IBM's document parser (PDF, DOCX, PPTX) with strong layout and table understanding; a default in the LlamaIndex/LangChain ecosystem (~64k stars).
- 5marker (Datalab)Fast, high-quality PDF/EPUB/DOCX to Markdown conversion with tables, math, and images; a general-purpose favorite (~38k stars).
Emerging
- 1Pythonation/Mistral-Arabic-OCR-testPython scripts that run Arabic PDFs and scans through Mistral's OCR API, returning Markdown that keeps tables and headings; batch mode included.
- 2mohcodes01/Document-Intelligence
- 3GramosoftAI/GdoczAISelf-hostable extraction platform where one sample document defines a schema, then Gemini, Mistral or Qwen OCR fills it and pushes JSON to ERPs.
- 4ComPDFKit/compdfkit-conversion-sdk-macCommercial Objective-C library for macOS apps that converts PDFs to Word, Excel, images and Markdown, with OCR and table recognition.
- 5EricRollei/PDF-ToolsComfyUI custom nodes for PDF page extraction, Surya and PaddleOCR text recognition, and Florence-2 layout and object analysis in visual workflows.
Skills
Skills you drop into a coding agent.Leading
- 1Anthropic pdf skill (official)The official Anthropic SKILL.md that powers Claude's PDF work — form filling, text/table extraction, merge, and generation. The reference Agent Skill for PDFs (in anthropics/skills, ~164k stars).
- 2claude-office-skills (tfriedel)Community office-document Agent Skills pack (PDF, DOCX, PPTX, XLSX) with automation helpers for Claude Code (~800 stars).
Emerging
- 1appautomaton/document-SKILLsAnthropic's PDF, Word, Excel and PowerPoint skills repackaged as a single plugin installable into Claude Code or Codex.
- 2deckflow/deckprobe
- 3PaodingAI/skillsThree agent skills calling PaodingAI's cloud API to convert PDFs, Office files and images to Markdown, diff two documents, and build ChatDOC knowledge bases.
- 4franciscojeg/ai-document-toolkit
- 5PSPDFKit-labs/nutrient-agent-skillInstalls into 40+ coding agents as a skill rather than an MCP server, calling Nutrient's cloud API to convert, OCR, redact and sign.