Docling

Docling is a document conversion library for turning varied files into structured data for AI workflows. It suits developers building document pipelines, especially when local processing matters. Its broad inputs, exports and deployment choices make it a strong free starting point; managed use brings service limits and provider-specific data terms.

Overview

Docling’s value is in preserving structure, not merely extracting text. Its PDF processing handles layout, reading order, tables, code, formulas and image classification, which can give downstream AI systems more context to work with. The project is hosted by the LF AI & Data Foundation and began with IBM Research Zurich’s AI for knowledge team.

It can run as a local Python library, a self-operated HTTP service, or through managed and private deployment offerings. Local processing can continue offline after models are downloaded and cached, without uploading documents. That makes the library a sensible choice for sensitive workloads that can be operated in-house. Managed services are less straightforward: requests face size, rate and concurrency limits, and provider terms determine retention and processing boundaries.

Key features

Broad input coverage: Supported formats include PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, audio, video, email and Apple iWork files. That range can consolidate varied sources into one conversion workflow, though the breadth alone does not establish identical handling across formats.

Structured exports: Targets include Markdown, HTML, JSON, DocLang XML, text, DocTags, WebVTT, DocLang archive, JSONL chunks and LaTeX. The mix supports both readable output and structured data for later processing.

OCR choices: Docling supports OCR through EasyOCR, Tesseract, OcrMac and RapidOCR. Multiple engines give teams options for their conversion pipeline rather than tying the workflow to one OCR engine.

AI framework integrations: Documented integrations include LangChain, LlamaIndex, Haystack and Crew AI, along with other frameworks and agent tools. This is useful for teams already assembling AI workflows around those ecosystems.

Batch and API conversion: API access, batch conversion and OCR conversion are supported, alongside local and self-operated deployment choices.

Support channels: The project links to a community Discord and a live assistant. Those offer routes to help, but they are not a stated service-level commitment.

Pricing

The Docling open-source library costs nothing and is MIT licensed. It is the clearest fit for teams able to install and operate the library themselves: local use and offline processing after model download avoid a managed-service quota, though they place operation with the user.

Docling for IBM watsonx is free for a 30-day trial, with 5,000 pages free. That gives teams a bounded way to evaluate the managed option; the page allowance and trial period make it less suitable as an assumed ongoing free tier. Managed service requests also have size, rate and concurrency limits, and provider terms govern retention and processing boundaries. Private deployment offerings are available with custom pricing.

The pricing model is free, with a free plan and free trial. The open-source library has no stated seat or page quota; the watsonx trial does have a 5,000-page allowance and lasts 30 days.

Platforms

Docling supports Windows, macOS and Linux, as well as API and self-hosted deployment. Local library use and self-operated service suit teams that need control over where documents are processed; managed deployment trades that control for provider-dependent limits and terms.

Who it's for

Docling is best suited to developers and teams building document-to-AI pipelines across mixed file types, especially when retaining layout and other structure matters. Its local and offline modes favor privacy-conscious workflows, while the API and framework integrations help teams connecting conversion to existing AI systems. It is a weaker fit for buyers who need a managed service without request limits or provider-dependent data terms, or who want a conventional end-user PDF editor rather than a conversion library.

Pros and cons

Pro — broad formats and targets: It accepts everything from office files and EPUB to audio, video and email, and exports to formats including Markdown, JSON, WebVTT and LaTeX.

Pro — local, offline operation: Once models are downloaded and cached, processing can run offline without uploading documents, a concrete advantage for sensitive material.

Pro — free MIT-licensed library: Teams can use the library at no cost and choose local or self-operated deployment.

Con — managed limits and terms: Size, rate and concurrency caps, plus provider-set retention and processing boundaries, require review before routing documents to a managed service.

Con — trial is bounded: The watsonx offer lasts 30 days and covers 5,000 pages, so it does not establish a continuing free managed tier.

Alternatives

Choose HiPDF instead if a freemium option with web access, OCR and batch processing across web, desktop, mobile and API platforms suits the job better; its Basic plan is $39.95 a year, billed annually.

HTML2PDF Rocket is a more focused alternative for HTML-to-PDF conversion; its free plan allows 200 credits per month and limits PDF size to 5 Mb, with large PDFs using one credit per 500kb of network traffic.

For a general conversion service with a stated monthly file allowance, consider Convertio: its Lite plan is $11.99 a month, cancel anytime, and includes a 1 GB maximum file size and 1,000 conversion credits per month.

FoxyUtils Online PDF Tools is a web-only alternative with OCR and cloud-connect features in its Monthly plan at $5.99 a month.

PDF Agile is another option for conversion; its free plan allows 2 pages and 5 conversion tasks.

Consider MinerU as another free open-source document-processing option.

Web to PDF is a free alternative for unlimited conversions without file or page limits, when converting web content to PDF is the requirement.

Ghostscript is another free option.

For broader category browsing, see PDF Converter Software and AI PDF Tools.

Verdict

Choose Docling if you need a no-cost, MIT-licensed foundation for structured document conversion, broad format coverage and local or offline AI workflows. Its strongest reason to choose is the combination of structural PDF processing and deployment control. Look elsewhere if you need a managed service with predictable uncapped requests or provider-independent data terms.