jztan/pdf-mcp

View on GitHub ↗

MCP server that lets Claude Code and other AI agents work through large PDFs, and whole folders of them, without overflowing context: hybrid semantic + keyword search, selective page reading, tables, images, OCR, chart data, and multi-column/CJK layouts.

114 ★8 forksPythonUpdated 18d ago

What you need to know

Surgical PDF access: semantic/keyword search, read-only page snippets, tables, images, scanned text

Install

pip install pdf-mcp

Usage

  • Query PDFs for relevant pages without sending whole documents

Key features

  • semantic and keyword search
  • read-only page extraction
  • table/image/scanned text support
  • multicolumn and Japanese text

Best for

Extracting facts from large PDFs cheaply

Caveats

  • Access is read-only by design
Platforms: PythonClients: Claude · any MCP client

Reviewed 2026-08-11

Topics

agentic-ragaicjkclaudeclaude-codecodex-clidocument-processingllmmcpmcp-servermodel-context-protocolocropencodepdfpdf-extractionpymupdfpythonragsemantic-searchtable-extraction
Stars
114★
Forks
8
Language
Python
License
MIT
Created
2026-01-28
Last push
2026-08-17