Docling

Docling is IBM’s open-source document parser that converts PDFs and Office files into structured JSON and markdown for AI apps.

Overview

Docling provides fast, accurate document conversion as an open-source library. It parses PDFs, DOCX, PPTX, XLSX, images, and more into a unified document format with layout, tables, and reading order preserved. Runs fully local — a decisive advantage for privacy-sensitive workloads — and integrates with LangChain, LlamaIndex, and popular vector stores. Its model zoo includes layout analysis and table recognition models that perform well on scientific and technical documents.

Developers like Docling for its engineering quality: fast inference, deterministic output schemas, and a document model that preserves hierarchy. IBM backs the project, and the community contributes format support and model improvements through the public repository.

Key Features

• Unified parsing of PDFs and Office formats
• Structured JSON and markdown output with layout fidelity
• Fully local processing for data privacy
• Table structure recognition and reading order
• Integrations with LangChain and LlamaIndex
• Free, open-source, and extensible model zoo

Docling’s document model is designed for downstream AI use: reading order, hierarchy, and table structure survive conversion. For organizations that cannot send documents to external APIs, it is often the only practical high-quality option.

Pricing

Docling is completely free and open source under an MIT-style license; there is no paid tier.

Who Is It For

Developers and enterprises that need document parsing without sending content to third-party clouds, especially in regulated industries.

FAQ

Is Docling free to use?

Yes, entirely free and open source.

Does Docling offer an API?

Docling is a Python library and CLI; it runs in your own environment rather than as a hosted API.

How does Docling compare to alternatives?

Hosted parsers add convenience and scale; Docling wins when data must stay local and costs must stay zero.

Does Docling need a GPU?

No, it runs on CPU as well, though a GPU accelerates large batch conversions.

What output formats are available?

Docling produces structured JSON and markdown suitable for RAG pipelines and archival workflows.

Does Docling support scanned PDFs?

Yes, OCR-backed parsing handles scans and images as well as born-digital files.

Who maintains Docling?

The project is developed open source with significant contributions from IBM Research.

Is Docling suitable for production pipelines?

Yes, deterministic output schemas make it reliable in automated ingestion workflows.

Does it support Office files?

Yes, DOCX, PPTX, and XLSX convert alongside PDFs and images.

Does Docling support markdown output?

Yes, markdown and structured JSON are the primary output formats, both designed for retrieval and language model consumption.