pdfmux — PDF extraction that actually works

Name: pdfmux — PDF extraction that actually works
Author: NameetP

@NameetP

About pdfmux — PDF extraction that actually works

MCP server for pdfmux — the only PDF extractor that audits every page. #2 on opendataloader-bench (200 real-world PDFs). 5 tools: get_pdf_metadata, convert_pdf, analyze_pdf, batch_convert, extract_structured. Zero GPU, zero API keys, zero cost. Give AI agents reliable PDF reading

Basic information

Config

Add this server to your MCP-compatible client using the configuration below.

{
  "mcpServers": {
    "pdfmux": {
      "command": "npx",
      "args": [
        "-y",
        "pdfmux-mcp"
      ]
    }
  }
}

Tools

No tools detected

We auto-extract tools from the README. The maintainer can list them under a ## Tools heading to populate this section.

Overview

What is pdfmux?

pdfmux is a universal PDF extraction orchestrator that routes each page to the best backend, audits output quality, and re-extracts failures automatically. It combines five rule-based extractors plus a bring-your-own-key LLM fallback in one CLI and Python API for developers and AI agents.

How to use pdfmux?

Install with pip install pdfmux, then run pdfmux convert file.pdf for zero-config extraction. Use extra install options like pdfmux[ocr] for scanned pages or pdfmux[llm] for LLM fallback. The Python API offers extract_text(), extract_json(), and chunk() for programmatic use.

Key features of pdfmux

Zero‑config extraction routes pages to best backend automatically
RAG chunking with configurable token limits per chunk
Cost modes (economy/balanced/premium) with budget caps
Schema-guided extraction with five built-in presets
BYOK LLM fallback supports Gemini, Claude, GPT-4o, Ollama
Built‑in MCP server for AI agent integration (stdio/HTTP)

Use cases of pdfmux

Extract text and tables from invoices for automated data entry
Chunk research papers into sections for RAG pipelines
Batch convert directories of PDFs to markdown or JSON
Extract structured data from scanned receipts with OCR fallback
Integrate with LangChain via langchain-pdfmux loader

FAQ from pdfmux

What backends does pdfmux use?

PyMuPDF (digital text), OpenDataLoader (complex layouts), RapidOCR (scanned pages), Docling (tables), Surya (heavy OCR), plus any BYOK LLM.

Does pdfmux require an LLM API key?

No — the LLM fallback is optional. The five rule‑based extractors work without any keys. LLM support requires a provider (Gemini, Claude, etc.) and its API key.