# pdfmux

A PDF extraction tool that checks its own output for dropped or scrambled pages and re-extracts them with a stronger backend, and can also audit any other extractor's output for the same silent failures.

pdfmux is a self-healing PDF extraction tool: rather than assuming an extraction succeeded, it audits its own output for blank pages, scrambled columns and broken tables, re-extracts those pages with a stronger of seven built-in backends or an optional bring-your-own-key vision model, and flags whatever it still cannot read instead of silently dropping it. A separate verification mode audits any extractor's output, such as another parser or service, against the source PDF and reports which pages it silently dropped.

It is used as a command-line tool, a Python API, an MCP server for AI assistants, or through LangChain and LlamaIndex loaders, and is offered as a free, open-source alternative to hosted PDF-parsing services for retrieval-augmented-generation pipelines. Handling scanned pages requires installing an additional OCR extra.

It requires Python 3.11 or newer and installs as a Python package.

Website: https://pdfmux.com