PDF MCP Server is a comprehensive PDF toolkit that enables AI-powered control of PDF operations through Claude via the Model Context Protocol (MCP). It runs as a FastMCP server exposing a suite of tools built around PDFtk and QPDF, allowing users to perform merge, split, encrypt, optimize, and analyze PDFs using natural language commands. The project supports Python 3.8+ and provides two installation approaches: via a virtual environment or a direct pip install of core dependencies. Claude Desktop integration is documented for Windows, macOS, and Linux, showing how to register the MCP server (with command and arguments) and how to restart Claude to begin issuing natural-language requests like “Merge these PDFs” or “Encrypt this document.” The repository lists core tool groups including core operations (merge_pdfs, split_pdf, extract_pages, rotate_pages), security (encrypt_pdf, decrypt_pdf), optimization/repair (optimize_pdf, repair_pdf, check_pdf_integrity), information/analysis (get_pdf_info, update_pdf_metadata, inspect_pdf_structure, extract_pdf_attachments), and file management. It also outlines architecture: server.py, wrappers, utilities, and dependencies: FastMCP, PDFtk, QPDF.