# Sezzlee PDF

> Sezzlee PDF is an open-source MCP server for files and documents. Read PDFs from one sandboxed folder as Markdown, with page search and optional OCR. It runs locally on your machine and is started by your AI client.

- **Kind:** MCP server
- **Pricing:** Open source
- **Platforms:** macOS, Windows, Linux, CLI
- **Website:** https://docs.sezzlee.app/docs/pdf-mcp/quickstart
- **Repository:** https://github.com/sezzlee/mcp/tree/HEAD/packages/servers/pdf-mcp
- **MCP Registry name:** io.github.sezzlee/pdf
- **Categories:** files and documents
- **Listing updated:** 2026-11-15

## About

This server reads, and only reads. Point it at a folder at startup and every PDF below it becomes available as page-numbered Markdown, with headings and tables kept as Markdown structure so an agent can read a revenue table by column. Nothing outside that folder is reachable: paths that climb out, absolute paths and symlinks leaving the root are refused, and error messages never reveal the folder's real location.

The usual sequence is three calls: list_documents to find a file, describe_document for page count, whether it is text or scanned and which pages lack readable text, then read_pages for just the pages needed. find_in_document does literal search and returns the page, line and surrounding text of each match. Pages with no text are flagged needsOcr instead of coming back blank.

OCR is off unless you pass a module with --ocr that supplies a rasterizer and a provider, so the connection decision stays with you; the server itself has no network code. It needs Node.js 22 or later on Apple-silicon macOS, glibc Linux or Windows x64, and does not run on Intel Macs or Alpine.

## Features

- **Folder sandbox:** Access is limited to one folder chosen at startup, including its subfolders.
- **Markdown pages:** Pages come back numbered from 1 with tables kept as Markdown.
- **Document summary:** describe_document reports page count, text or scanned, and pages that need OCR.
- **Literal search:** find_in_document returns the page, line and context of each match.
- **Optional pluggable OCR:** Scanned pages are read only when an OCR module is supplied with --ocr.
- **No network access:** The server has no network code and no write tools.

## Good for

- Ask which region had the most revenue in an annual report
- Search a folder of contracts for a clause by literal text
- Check which pages of a scan need OCR
- Pull one table out of a long report without loading the whole file

## Install

### Claude Code (Terminal)

```sh
claude mcp add pdf -- npx -y @sezzlee/pdf-mcp
```

### Claude Desktop (claude_desktop_config.json)

```json
{
  "mcpServers": {
    "pdf": {
      "command": "npx",
      "args": [
        "-y",
        "@sezzlee/pdf-mcp"
      ]
    }
  }
}
```

### Cursor (~/.cursor/mcp.json)

```json
{
  "mcpServers": {
    "pdf": {
      "command": "npx",
      "args": [
        "-y",
        "@sezzlee/pdf-mcp"
      ]
    }
  }
}
```

### VS Code (.vscode/mcp.json)

```json
{
  "servers": {
    "pdf": {
      "type": "stdio",
      "command": "npx",
      "args": [
        "-y",
        "@sezzlee/pdf-mcp"
      ]
    }
  }
}
```

### Windsurf (~/.codeium/windsurf/mcp_config.json)

```json
{
  "mcpServers": {
    "pdf": {
      "command": "npx",
      "args": [
        "-y",
        "@sezzlee/pdf-mcp"
      ]
    }
  }
}
```

### Other clients (your client's MCP settings (most use this shape))

```json
{
  "mcpServers": {
    "pdf": {
      "command": "npx",
      "args": [
        "-y",
        "@sezzlee/pdf-mcp"
      ]
    }
  }
}
```

## Tools

- `list_documents`: List readable PDF candidates under the root
- `describe_document`: Page count, document type and per-page OCR need
- `read_pages`: Read selected pages as Markdown
- `find_in_document`: Literal search over extracted text with page and context

## FAQ

### Can it change my PDFs?

No. It has no write tools, so documents are never modified.

### How do I limit what it can see?

Pass one absolute folder path as the only argument; everything outside is refused.

### Does it do OCR?

Only if you start it with --ocr and a module that supplies a rasterizer and an OCR provider.

### Where does it run?

On Apple-silicon macOS, glibc Linux and Windows x64 with Node.js 22 or later.

### Is Sezzlee PDF free?

Yes. Sezzlee PDF is open source, so you can use and self-host it for free.

### How do I add Sezzlee PDF to Claude Code?

Run this command in your terminal: claude mcp add pdf -- npx -y @sezzlee/pdf-mcp

### What are the best alternatives to Sezzlee PDF?

The closest alternatives to Sezzlee PDF are ASTra, CARTESIAN ECU, Invoices, SigPen and ConvertAudioToText.

## Alternatives

- [ASTra](https://ai-powered.com/mcp/astra): ASTra is a freemium MCP server for files and documents.
- [CARTESIAN ECU](https://ai-powered.com/mcp/cartesian-ecu): CARTESIAN ECU is a paid MCP server for files and documents.
- [Invoices](https://ai-powered.com/mcp/invoices): Invoices is a paid MCP server for files and documents.
- [SigPen](https://ai-powered.com/mcp/sigpen): SigPen is a freemium MCP server for files and documents.
- [ConvertAudioToText](https://ai-powered.com/mcp/convertaudiototext): ConvertAudioToText is a freemium MCP server for design and media.
- [WhoScribe](https://ai-powered.com/mcp/whoscribe): WhoScribe is a freemium MCP server for design and media.
- [filex](https://ai-powered.com/mcp/filex): filex is an open-source MCP server for files and documents.
- [DocJacket](https://ai-powered.com/mcp/docjacket): DocJacket is a freemium MCP server for productivity.

---
Source: https://ai-powered.com/mcp/sezzlee-pdf
