PDF MCP

What it does

PDF MCP provides AI agents with precise access to PDF documents. Instead of flooding the context window with raw text, it allows agents to search, skim, and extract specific pages, tables, and images from large PDFs using a combination of keyword and semantic search.

Tools

• pdf_info: Get page count, metadata, and TOC summary.
• pdf_search: Perform hybrid RRF search (keyword + semantic) to find relevant sections.
• pdf_read_pages: Extract text, tables, and images from specific page ranges with optional OCR.
• pdf_render_pages: Render pages as PNGs for vision models.
• pdf_read_all: Read the entire document (byte-capped).
• pdf_get_toc: Retrieve full table of contents.
• pdf_cache_stats: View document cache usage.
• pdf_cache_clear: Clear cached document data.
• server_info: Check active optional features like OCR or semantic search.

Installation

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

Supported hosts

Claude Desktop, Claude Code, VS Code Copilot, Codex CLI, Kiro