Skip to the content
Know your tech before you buy it. Makers: get listed from $49/mo

Docling

docling-project.github.io

  • Free plan
  • Windows
  • macOS
  • Linux
  • API
  • Self-hosted

What Docling is

Docling converts documents into structured data for AI workflows. It accepts PDFs, Office documents, HTML, EPUB, audio, video, images, email, and Apple iWork files. For PDFs, it handles page layout and reading order, as well as tables, code, formulas, and image classification. Output options include Markdown, HTML, WebVTT, DocLang, DocTags, lossless JSON, text, LaTeX, and JSONL chunks. The project supports multiple OCR engines, including EasyOCR, Tesseract, OcrMac, and RapidOCR, and documents integrations with LangChain, LlamaIndex, Haystack, Crew AI, and other AI frameworks and agent tools. Docling can run as a local Python library, a self-operated HTTP service, or through managed and private deployment offerings. The project says local use can continue offline after models are downloaded and cached, without uploading documents. Managed requests have size, rate, and concurrency limits, and provider terms define retention and processing boundaries. The open-source library is free and MIT licensed; a separate IBM watsonx trial is free for 30 days and covers 5,000 pages. The project is hosted by the LF AI & Data Foundation and was started by IBM Research Zurich's AI for knowledge team.

Who it suits

Docling suits developers and AI teams converting varied documents into structured outputs for downstream workflows. Its local offline mode may suit users who want to process documents without uploading them.

Good to know, and what to watch

  • Accepts documents, media, email, and iWork files.
  • Handles PDF layout, tables, code, and formulas.
  • Supports several OCR engines.
  • Can run locally and offline after model downloads.
  • Open-source library is free and MIT licensed.
  • Managed requests have size, rate, and concurrency limits.
  • Provider terms determine managed-service retention and processing boundaries.
  • IBM watsonx trial is limited to 5,000 pages and 30 days.

Docling for IBM watsonx trial

Free

Docling open-source library

Free

Open source

Getting started

  1. Open the Docling project website
  2. Install the MIT-licensed library locally or run a self-operated HTTP service
  3. Download and cache the models for offline processing
  4. Connect Docling to an AI framework or agent tool if needed
  5. Use the IBM watsonx offer for its free 30-day, 5,000-page trial

Limits to check

The IBM watsonx trial is free for 30 days and covers 5,000 pages. Managed service requests are subject to size, rate, and concurrency limits, with provider terms determining retention and processing boundaries.

The Club's review

Docling is an open-source document conversion project that turns varied files into structured data for AI workflows. It is best suited to developers and teams building document pipelines, especially when local processing matters. Its strength is the combination of broad format support, structured PDF handling, and offline use; its managed options come with operational limits and provider-dependent data terms.

Overview

Docling is designed to preserve more than a document’s words. Its PDF processing covers page layout, reading order, tables, code, formulas, and image classification, which can help downstream systems interpret how content is organized. It can run as a local Python library, a self-operated HTTP service, or through managed and private deployment offerings. The project is hosted by the LF AI & Data Foundation and began with IBM Research Zurich’s AI for knowledge team.

Local execution can continue offline after models are downloaded and cached, without uploading documents. That makes the library a strong fit for sensitive pipelines that can be operated in-house. Managed services add convenience, but requests face size, rate, and concurrency limits, and provider terms govern retention and processing boundaries.

Key features

  • Wide input and output coverage: Supported inputs include PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, audio, video, email, and Apple iWork files. Exports include Markdown, HTML, WebVTT, DocLang, DocTags, and lossless JSON; conversion targets also include plain text, DocLang XML and archives, chunks in JSONL, and LaTeX. This range suits pipelines that need to normalize different source materials for AI use.
  • Structured PDF processing: Layout, reading order, table structure, code, formulas, and image classification give Docling a broader document understanding role than simple text extraction. Teams needing those relationships carried into later processing have a clear reason to consider it.
  • OCR choices: EasyOCR, Tesseract, OcrMac, and RapidOCR are supported. Multiple engines offer options for OCR workflows, alongside batch conversion.
  • AI framework integrations: Documented integrations include LangChain, LlamaIndex, Haystack, and Crew AI, as well as other frameworks and agent tools. This helps teams connect conversion to an existing AI stack.
  • Local and service deployment: Run the library locally, operate it as an HTTP service, or use managed and private deployment offerings. The flexibility supports different operating models, while managed use requires attention to provider limits and terms.
  • Community support: A community Discord and live assistant provide support channels for users.

Pricing

The open-source library costs nothing and is MIT licensed. It supports local installation and offline processing after model download, making it the natural choice for teams that can manage their own environment and want to avoid a subscription for the library.

Docling for IBM watsonx is free for a 30-day trial, with free processing for 5,000 pages. That gives teams a defined way to evaluate the managed offering, but the trial is time-limited and the page allowance is capped. Managed service requests also face size, rate, and concurrency limits; provider terms determine retention and processing boundaries.

The project’s pricing model is free, with a free plan and trial. The library is the most direct option for ongoing use when self-operation is suitable; the watsonx offer is a bounded trial rather than an equivalent unlimited hosted plan.

Platforms

Docling supports API access and runs on Linux, macOS, and Windows, with self-hosted deployment also supported. That makes it viable for teams building server-side pipelines or running conversions within their own infrastructure, as well as developers working across the major desktop operating systems.

Who it's for

Docling is a good fit for developers, AI teams, and IT groups that need to convert mixed document collections into structured outputs, especially when PDF layout and local processing matter. Its range of inputs and AI integrations suits teams building repeatable document workflows. Organizations that prefer a managed service should weigh the request caps and provider-specific data terms; readers looking mainly for a simple PDF utility may find a narrower tool more appropriate.

Pros and cons

  • Pro: The free MIT-licensed library supports local and offline use after model downloads, giving teams control over where document processing happens.
  • Pro: Broad input formats, structured PDF handling, multiple OCR engines, and several export targets make it useful for varied AI document pipelines.
  • Pro: Local library, self-operated HTTP service, and managed or private deployment options accommodate different operating models.
  • Con: Managed service requests are subject to size, rate, and concurrency limits, which can constrain production workloads.
  • Con: Provider terms define retention and processing boundaries for managed use, so teams with strict data requirements must evaluate those terms.
  • Con: The IBM watsonx offer is limited to a 30-day trial and 5,000 free pages, so it does not provide an open-ended hosted allowance.

Alternatives

For a general PDF tool, PDF Converter Software is a broader category to compare. Readers focused on AI-assisted PDF workflows can also browse AI PDF Tools.

Choose HiPDF if you want a freemium PDF service with web and mobile access as well as OCR and batch processing; its Basic plan is $39.95 a year. FoxyUtils Online PDF Tools is a web-based alternative with a $5.99 a month plan that includes OCR, batch processing, and unlimited access.

PDFCreator is an alternative for Windows and self-hosted PDF work; it is freemium and open source. Pick pdflayer when API-based PDF conversion is the priority: its free plan allows 100 API requests per month for non-commercial use, and its Basic plan is $9.99 a month.

HTML2PDF Rocket is another freemium option, with API access and a free allowance of 200 credits per month, subject to file-size and traffic-based credit rules. PDF Agile offers a free plan capped at 2 pages and 5 conversion tasks, which may suit occasional small jobs. Consider MinerU as another free, self-hosted document processing project. Web to PDF suits straightforward web-to-PDF conversion, with unlimited conversions and no file or page limit on its free plan.

Verdict

Docling is the right choice for developers and teams building AI document pipelines that need broad format support, structured PDF extraction, and local or offline processing. Its free MIT-licensed library gives those teams substantial flexibility without a software fee. Look elsewhere if your priority is a managed PDF service with predictable request capacity and clear data terms without having to evaluate provider limits.

The verdict

Docling offers broad input and export support, several OCR choices, and local processing options for AI workflows. Teams considering managed deployment should review its request limits and provider terms.

Questions people ask

Is Docling free?

Yes. The Docling open-source library costs nothing and is MIT licensed. The IBM watsonx offer is also free for a 30-day trial covering 5,000 pages.

Which inputs can Docling convert?

Supported inputs include PDF, DOCX, PPTX, XLSX, HTML, EPUB, audio, video, images, email, and Apple iWork files.

Can Docling run offline?

Yes. The project says local use can run offline after its models have been downloaded and cached, without uploading documents.

What output formats are available?

Outputs include Markdown, HTML, WebVTT, DocLang, DocTags, lossless JSON, text, LaTeX, and JSONL chunks, among other documented targets.

Where can Docling run?

Docling is available as a local Python library, a self-operated HTTP service, and managed or private deployments. Listed platforms include Windows, macOS, Linux, API, and self-hosted use.

What the maker publishes

AI integrations
Documented integrations include LangChain, LlamaIndex, Haystack, Crew AI, and other AI frameworks and agent tools. source
Deployment options
Docling can run as a local Python library, as a self-operated HTTP service, or through managed and private deployment offerings. source
Exports
Docling can export to formats including Markdown, HTML, WebVTT, DocLang, DocTags, and lossless JSON. source
Local privacy
The project says it can run locally, including offline after its models are downloaded and cached, without uploading documents. source
Managed service limits
The deployment page says managed service requests have size, rate, and concurrency limits, and provider terms determine retention and processing boundaries. source
OCR
The project supports multiple OCR engines, including EasyOCR, Tesseract, OcrMac, and RapidOCR. source
PDF understanding
Its PDF processing handles page layout, reading order, table structure, code, formulas, and image classification. source
Platforms
The installation documentation says Docling works on macOS, Linux, and Windows on x86_64 and arm64 architectures. source
Project home
Docling is hosted as a project in the LF AI & Data Foundation and was started by IBM Research Zurich's AI for knowledge team. source
Purpose
Docling converts documents such as PDFs, Office files, HTML, images, and audio into structured data for AI workflows. source
Support
The documentation links to a Docling community Discord and a live assistant for support. source
Supported inputs
The documentation lists formats including PDF, DOCX, PPTX, XLSX, HTML, EPUB, audio, video, images, email, and Apple iWork files. source

Checked against the maker's pages on 5 Oct 2026.