pdf-mcp

Overview

pdf-mcp is a specialized MCP server built with Python and PyMuPDF that provides AI agents with efficient access to PDF content. It solves the context window limitation problem by allowing agents to read specific pages or ranges rather than loading entire documents. The server implements hybrid search combining BM25 keyword and semantic search via Reciprocal Rank Fusion, OCR capabilities for scanned documents, structured extraction of tables and images, and persistent SQLite-based caching. It features robust security with HTTPS-only URL fetching and SSRF protection.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• pdf_info

  Page count, metadata, TOC summary, scanned-page detection. Call first.

  Try:What information can you tell me about this PDF?Try:How many pages are in the document?
• pdf_get_toc

  Full table of contents for documents with >50 bookmarks

  Try:Show me the table of contentsTry:What are the main sections of this document?
• pdf_read_pages

  Read specific pages or ranges; OCR-on-demand; embedded images + tables

  Try:Read pages 10-15 of the documentTry:OCR pages 3-5 of the scanned PDF
• pdf_read_all

  Read entire document in one call (byte-capped for safety)

  Try:Read the entire documentTry:Summarize this PDF for me
• pdf_render_pages

  Render pages as PNG for vision models — diagrams, handwriting, scans

  Try:Show me what page 5 looks likeTry:Render pages 10-15 as images
• pdf_search

  Hybrid RRF search (keyword + semantic), page or section granularity

  Try:Find sections about the training processTry:Which pages discuss supply chain risks?
• pdf_cache_stats

  Per-document cache breakdown + total size

  Try:Show me cache statisticsTry:How much space is used for caching?
• pdf_cache_clear

  Clear expired or all cache entries

  Try:Clear the cacheTry:Remove all cached data

Comparable tools

Installation

pip install pdf-mcp

{
  "mcpServers": {
    "pdf-mcp": {
      "command": "pdf-mcp"
    }
  }
}

FAQ

How does pdf-mcp handle large PDFs?

pdf-mcp uses chunked reading, allowing AI agents to read specific pages or page ranges rather than loading entire documents, preventing context overflow issues.

What search capabilities are available?

pdf-mcp provides hybrid search using Reciprocal Rank Fusion to combine BM25 keyword search and semantic search, enabling more comprehensive document queries.

Compare pdf-mcp with