
How to add PDF Reader MCP to Claude Desktop
Extract text, images, and metadata from local or URL PDFs as structured output, dir-confined. Paste the config into ~/Library/Application Support/Claude/claude_desktop_config.json and restart Claude Desktop.
Last updated June 15, 2026 ยท 200โ ยท stdio ยท no auth
Claude Desktop config for PDF Reader MCP
npx -y @sylphx/pdf-reader-mcp{
"mcpServers": {
"pdf-reader-mcp": {
"command": "npx",
"args": [
"-y",
"@sylphx/pdf-reader-mcp"
]
}
}
}Setup steps
- 1Open Claude Desktop โ Settings โ Developer โ Edit Config (this opens ~/Library/Application Support/Claude/claude_desktop_config.json).
- 2Paste the PDF Reader MCP config below under the top-level "mcpServers" key.
- 3Fill in any placeholder secrets (API keys, paths) in the snippet.
- 4Save the file, then fully quit and reopen Claude Desktop.
- 5Open a chat and confirm PDF Reader MCP's tools appear under the ๐ tools menu.
Before you start
- Node.js 22.13.0 or higher (required by pdfjs-dist v6)
- npx (bundled with Node.js) to run @sylphx/pdf-reader-mcp without a global install
- No credentials or API key needed โ auth is none
- Optional: a directory to confine reads to (passed via --allow-dir), and network access if reading PDFs from URLs
What PDF Reader MCP can do in Claude Desktop
read_pdfUnified tool to extract text, images, metadata, and page count from one or more PDF sources (local paths or URLs), with optional per-source page-range selection.
Security
File access is confined to the working directory the host sets, so run it with the cwd scoped to the intended project folder. Parsing untrusted PDFs always carries some parser-exploitation risk; only process documents you trust. Note: use the current @sylphx/ package, not the older @sylphlab/ name.
PDF Reader MCP + Claude Desktop FAQ
Where is the Claude Desktop config file?
Claude Desktop reads MCP servers from ~/Library/Application Support/Claude/claude_desktop_config.json. Paste the PDF Reader MCP config there under the "mcpServers" key and restart the client.
Is PDF Reader MCP safe to use with Claude Desktop?
File access is confined to the working directory the host sets, so run it with the cwd scoped to the intended project folder. Parsing untrusted PDFs always carries some parser-exploitation risk; only process documents you trust. Note: use the current @sylphx/ package, not the older @sylphlab/ name.
Does it need an API key or authentication?
No. It runs locally over stdio with no auth, reading PDFs from the filesystem or from HTTP/HTTPS URLs you allow.
Can it read PDFs from a URL, not just local files?
Yes. Each source can specify a `url` field for HTTP/HTTPS PDFs. You can disable this with `--no-http` or restrict it to specific domains with `--allow-host`.
How do I restrict which directories it can read?
Pass `--allow-dir=/path` (repeatable) or set the `MCP_PDF_ALLOWED_DIRS` environment variable. Reads outside allowed directories fail fast with an Access denied error.