PDF Reader MCP

Content

PDF Reader MCP is a dedicated Model Context Protocol server for extracting content from PDFs. It exposes a read_pdf tool that can process multiple local files or remote URLs, select page ranges, return metadata and page counts, extract full text, include embedded images, and detect table-like structures.

The project is focused on PDF reading rather than broad document conversion. It is useful when Claude needs page-aware PDF evidence while keeping the extraction surface limited to PDF files.

Source Review

• https://github.com/SylphxAI/pdf-reader-mcp
• https://github.com/SylphxAI/pdf-reader-mcp/blob/main/README.md
• https://registry.npmjs.org/%40sylphx%2Fpdf-reader-mcp
• https://github.com/SylphxAI/pdf-reader-mcp/blob/main/package.json
• https://github.com/SylphxAI/pdf-reader-mcp/blob/main/src/index.ts
• https://github.com/SylphxAI/pdf-reader-mcp/blob/main/src/handlers/readPdf.ts
• https://github.com/SylphxAI/pdf-reader-mcp/blob/main/src/pdf/loader.ts
• https://github.com/SylphxAI/pdf-reader-mcp/blob/main/src/utils/config.ts
• https://github.com/SylphxAI/pdf-reader-mcp/blob/main/LICENSE

These sources were reviewed on 2026-06-06. Prefer the live repository, README, npm registry metadata, package metadata, server entrypoint, read_pdf handler, PDF loader, security configuration source, and license file for current install commands, source restrictions, extraction behavior, and licensing.

Features

• npm package @sylphx/pdf-reader-mcp.
• Stdio MCP server launched with npx -y @sylphx/pdf-reader-mcp.
• Single read_pdf tool for one or more sources.
• Local path and remote URL sources.
• Page-range extraction with numbers or range strings.
• Optional full text, metadata, page count, image, and table extraction.
• Batched source and page processing to limit memory pressure.
• Directory allowlists, host allowlists, URL disable flag, and private-IP SSRF guard.
• MIT license.

Installation

Configure the stdio server in your MCP client:

{
  "mcpServers": {
    "pdf-reader": {
      "command": "npx",
      "args": ["-y", "@sylphx/pdf-reader-mcp"]
    }
  }
}

Replace the example directory and host with locations you trust before use. Keep MCP_PDF_ALLOW_HTTP set to false unless you have a specific, trusted plaintext HTTP source and understand the tampering risk. After restarting the MCP client, ask Claude to read only approved PDF paths or URLs and specify whether you need full text, metadata, page counts, images, tables, or particular page ranges.

Use Cases

• Extract text from a specific page range in a PDF.
• Read metadata and page count before deciding whether to process a document.
• Pull embedded images from selected PDF pages.
• Extract table-like content for review.
• Batch several approved PDFs in one tool call.
• Summarize a report, contract, manual, or invoice from extracted text.
• Restrict the server to approved directories or remote hosts for document workflows.

Safety and Privacy

PDF Reader MCP can expose everything a PDF contains, including hidden metadata, embedded images, form fields, author information, comments, and text that was not obvious from a quick visual scan. Use directory and host allowlists for private workflows, and verify extraction quality before relying on important tables, page ranges, or scanned-document text.

Remote URL fetching is enabled by default in the reviewed source, with configuration available to disable it or restrict hosts. Treat remote PDF URLs, local paths, extracted text, image data, tables, and metadata as sensitive unless the document is approved for the model session.

Duplicate Check

Existing content includes MarkItDown, Kreuzberg, Markdownify, and research servers that can process PDFs among many other formats. This entry is distinct because it covers SylphxAI/pdf-reader-mcp, a dedicated PDF-only MCP server with page ranges, images, tables, metadata, local/URL sources, and configurable file/URL restrictions. No matching source URL or dedicated PDF Reader MCP entry was found in content/mcp.

• Source repositorygithub.com 2026-07-21T07:17:01+00:00
• Documentationgithub.com
• Submitted by oktofeesh12026-06-06