How to Give Claude PDF Access via MCP Server

Mar 4, 2026

I wanted Claude to read PDFs on my local machine without uploading them anywhere. No cloud processing, no external services. Just local access to my documents. Turns out there’s a solution using MCP (Model Context Protocol) and pdf-oxide-mcp.

The Problem

Claude has excellent text capabilities but can’t directly read PDF files. When you need to extract text from a report, search through a contract, or convert a paper to markdown, you’re stuck. You could upload the PDF to cloud services, but that’s a non-starter for sensitive documents.

Enter MCP - Anthropic’s open standard for AI assistants to communicate with tools. MCP servers run locally and expose capabilities to Claude through a standardized interface. No cloud upload required.

Installing pdf-oxide-mcp

You have three options for installation:

# Homebrew (macOS/Linux)
brew install yfedoseev/tap/pdf-oxide

# Cargo
cargo install pdf_oxide_mcp

# crgx (recommended for Claude)
crgx pdf_oxide_mcp@latest

I recommend crgx for Claude Desktop users since it manages dependencies cleanly.

Configuration

Add the pdf-oxide server to your Claude configuration file:

{
  "mcpServers": {
    "pdf-oxide": {
      "command": "crgx",
      "args": ["pdf_oxide_mcp@latest"]
    }
  }
}

Config file locations:

Claude Desktop (macOS): ~/Library/Application Support/Claude/claude_desktop_config.json
Claude Desktop (Windows): %APPDATA%\Claude\claude_desktop_config.json
Claude Code: ~/.claude.json or .claude/settings.json

Available Tools

Once configured, you get these tools:

Tool	Description
`extract_text`	Extract plain text from a PDF
`search_text`	Search PDF content with regex
`convert_to_markdown`	Convert PDF to Markdown format

Example Workflows

After setup, try these prompts in Claude:

Extract and summarize:

“Extract the text from report.pdf and summarize the key findings.”

Convert to Markdown:

“Convert paper.pdf to markdown format, preserving the heading structure.”

Search within PDFs:

“Search contract.pdf for any mentions of ‘termination’ and summarize those sections.”

Privacy Architecture

The key benefit is local processing:

Claude          pdf-oxide-mcp          Local PDFs
  │                   │                    │
  │    (local call)   │   (file access)   │
  │──────────────────▶│──────────────────▶│
  │                   │                    │
  │   (no external)   │                    │
  │   data transfer   │                    │

Your PDFs never leave your machine. All processing happens locally through the MCP server.

When This Works

You have PDFs on your local machine
Claude Code or Claude Desktop is installed
You can install the MCP server

When This Doesn’t Work

PDFs are in cloud storage (download first)
You need OCR for scanned documents
PDFs are heavily encrypted

Summary

In this post, I showed how to give Claude local PDF access via MCP. The key point is privacy-first local processing without cloud uploads.

Final Words + More Resources

My intention with this article was to help others share my knowledge and experience. If you want to contact me, you can contact by email: Email me

Here are also the most important links from this article along with some further resources that will help you in this scope:

👨‍💻 pdf-oxide-mcp GitHub
👨‍💻 MCP Documentation

Oh, and if you found these resources useful, don’t forget to support me by starring the repo on GitHub!