PDF MCP Server for Claude, ChatGPT and Cursor

Install the PDF MCP server

You need Node.js 22 or newer. The first run downloads the server through npx. The folders after the package name are the only places it may read and write, so change them to wherever your PDFs live.

{
  "mcpServers": {
    "utilitly": {
      "command": "npx",
      "args": ["-y", "@utilitly/mcp", "~/Documents", "~/Downloads", "~/Desktop"]
    }
  }
}

{
  "mcpServers": {
    "utilitly": {
      "command": "npx",
      "args": ["-y", "@utilitly/mcp", "~/Documents", "~/Downloads", "~/Desktop"]
    }
  }
}

The server is also listed in the official MCP Registry as com.utilitly/mcp, so apps and directories that read the registry can find it by that name.

Claude, ChatGPT, Codex, Gemini and Cursor are trademarks of their owners and appear here only to show which apps the server works with. Utilitly isn't affiliated with or endorsed by them.

What your AI can do with it

Eight tools, each named so the AI app can tell what it does. Results land in the same folder as the input, and a name that's already taken gets -2, -3 and so on, so a file is never replaced.

Things to ask

Once it's installed, ask in plain words and use full paths:

• “Merge ~/Downloads/march.pdf and ~/Downloads/april.pdf.”
• “Delete the blank page 3 from ~/Desktop/scan.pdf.”
• “Split ~/Documents/handbook.pdf into pages 1-12, 13-30 and 31 to the end.”
• “Pages 2 to 4 of ~/Desktop/scan.pdf are sideways; turn them the right way up.”
• “How many pages does ~/Documents/contract.pdf have, and is it locked?”

What it can and can't touch

• It never overwrites. Input files are only read, never changed.
• With folders listed at install, nothing outside them can be read or written. The check runs after symlinks are resolved.
• Paths must be absolute or start with ~/, so the AI can't wander into the wrong folder through a relative path.
• Password-protected PDFs are refused rather than altered.
• Merging and other rebuilds don't carry over bookmarks or working form fields. Rotating keeps the whole document.

Open-source PDF libraries for JavaScript

The same engines also ship as six npm packages for your own code. Each one takes a PDF as bytes and returns new bytes, so the same code runs in Node.js and in the browser. They ship ESM and CommonJS builds with TypeScript types, and their one dependency is pdf-lib.

import { readFile, writeFile } from 'node:fs/promises'
import { mergePdfs } from '@utilitly/pdf-merge'

const book = await mergePdfs([
  await readFile('cover.pdf'),
  await readFile('chapter-1.pdf'),
  await readFile('chapter-2.pdf'),
])
await writeFile('book.pdf', book)

import { readFile, writeFile } from 'node:fs/promises'
import { splitPdf } from '@utilitly/pdf-split'

const parts = await splitPdf(await readFile('report.pdf'), '1-2, 3, 4-end')
for (const [i, part] of parts.entries()) {
  await writeFile(`report-part${i + 1}.pdf`, part)
}

import { readFile, writeFile } from 'node:fs/promises'
import { rotatePdf } from '@utilitly/pdf-rotate'

const fixed = await rotatePdf(await readFile('scan.pdf'), 90, { pages: '2, 5-end' })
await writeFile('scan-fixed.pdf', fixed)

import { readFile, writeFile } from 'node:fs/promises'
import { deletePages } from '@utilitly/pdf-delete-pages'

const trimmed = await deletePages(await readFile('scan.pdf'), '2, 5-end')
await writeFile('scan-trimmed.pdf', trimmed)

import { readFile, writeFile } from 'node:fs/promises'
import { reorderPages } from '@utilitly/pdf-reorder-pages'

const deck = await reorderPages(await readFile('deck.pdf'), [6, 1, 2, 3, 4, 5])
await writeFile('deck-summary-first.pdf', deck)

import { readFile, writeFile } from 'node:fs/promises'
import { duplicatePages } from '@utilitly/pdf-duplicate-pages'

const sheets = await duplicatePages(await readFile('form.pdf'), '2', { copies: 3 })
await writeFile('form-with-copies.pdf', sheets)

Questions