Docling (IBM)
Tool : A precise action the AI can call · requires opensource
Local PDF and document parsing, tables and layout included.
The best free option: everything stays on your machine.
Security reflects where data goes, requested permissions and incidents on record at the verification date. It is not a guarantee of zero risk: check the deployment mode and the connections you enable.
What it does, concretely
Extracting structured text from a PDF or office document is often tedious: layout is lost, tables become unreadable, and formulas disappear. Docling solves this by deeply analysing document structure, including reading order, tables, formulas, and image classification.
In practice, you install the Python library, pass it a local file or a URL, and it returns a structured document exportable as Markdown, HTML, or JSON. The tool runs entirely locally: no data leaves your machine. It supports a wide range of formats including PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, audio, video, emails, and more.
Docling is aimed at developers and teams building document processing pipelines, retrieval-augmented generation (RAG) systems, or AI agents. It integrates natively with LangChain, LlamaIndex, Crew AI, and Haystack, and also exposes an MCP server to connect any agent.
However, it is not suited for people without Python knowledge: there is no graphical interface, and everything goes through the command line or code. Users looking for a ready-to-use tool with a visual interface will need to look elsewhere.
What it offers
- ·Advanced PDF understanding: page layout, reading order, table structure, formulas, and image classification.
- ·Support for many input formats: PDF, DOCX, PPTX, XLSX, HTML, EPUB, images, audio (WAV, MP3), video (MP4, AVI, MOV, MKV, WebM), emails (EML, MSG), Apple Pages, LaTeX, and more.
- ·Export to multiple output formats: Markdown, HTML, lossless JSON, WebVTT, and DocTags.
- ·Extensive OCR for scanned PDFs and images, with support for several visual language models including GraniteDocling.
- ·Native integrations with LangChain, LlamaIndex, Crew AI, and Haystack for building AI pipelines.
- ·MCP server to connect Docling to any agent, and API server mode (docling-serve) to expose it as a service.
- ·Chart understanding (bar charts, pie charts, line plots): conversion to tables or code with detailed descriptions.
- ·Fully local execution, suitable for air-gapped environments and sensitive data.
In practice
- Price
- Free
- Commercial use
- Commercial use: yes
- Available on
- sdk-python, langchain
- Host
- opensource
- Licence
- MIT
- FR interface
- No
Security and data
- Rating
- Security: nothing to flag
- Capabilities
- readswritesexecutesnetwork
- Your data
- Data stays local
- Maintenance
- Active
- Last release
- 23 October 2026
Links and repository checked on 25 October 2026 · added on 1 July 2024
How to install Docling (IBM)
pip install docling