n8n-nodes-pdf (PDF Toolkit)

This is an n8n community node for working with PDFs directly inside your
workflows: merge, split, generate, fill forms, watermark, and extract.
Everything runs in-process on your n8n instance: no external API, no
per-document fees, and no data leaving your machine.

n8n is a fair-code licensed workflow automation platform.

Installation ·
Operations ·
Example usage ·
Limits ·
Credentials ·
Compatibility ·
Resources ·
Version history

Installation

Follow the installation guide
in the n8n community nodes documentation: Settings → Community Nodes →
Install, then enter n8n-nodes-pdf. No credentials or extra setup needed.

The package has zero runtime dependencies. The PDF engine ships bundled
inside the node.

Operations

One node, PDF Toolkit, with five resources. Operations read their input
PDF from a binary property (default data) and write results to a binary
property configurable under Options.

Document

Generate

Generation, Stamp → Text Watermark/Page Numbers, and Form → Fill Form use
bundled Unicode fonts (Noto Sans/Noto Sans Mono, plus Noto Emoji as a
monochrome fallback for emoji/pictographic characters), so full Latin, Latin
Extended (e.g. Polish "ł", Turkish "ı", Czech "č"), Cyrillic, and Greek text
is supported, not just the ASCII/WinAnsi range earlier versions were limited
to. Emoji render in monochrome only (no color), and a ZWJ-joined emoji
sequence (e.g. a "family" emoji built from several codepoints) may render as
its separate component emoji if the bundled font has no combined glyph for
that exact sequence. A character neither font covers (rare; mostly Private
Use Area codepoints, or scripts like CJK/Arabic/Hebrew that need a different
font family) throws a clear error naming the character instead of silently
dropping it or crashing with a raw font-library error. Custom fonts, nested
lists, and per-column table widths are still not supported. See
THIRD_PARTY_NOTICES.md for the bundled fonts'
licenses (all OFL-1.1).

Form

Stamp

Extract

Embedded Images extracts the raw image streams without re-encoding, which
works for JPEG-compressed images, the common case for photos and scans.
Images stored with other compression filters raise a clear error naming the
unsupported filter.

Example usage

Set these parameters on a PDF Toolkit node. Point Binary Property at
whatever binary field the previous node produced (usually data).

Merge all incoming items into one PDF:

{
  "resource": "document",
  "operation": "merge",
  "mergeFrom": "allItems",
  "binaryPropertyName": "data",
  "options": { "outputFileName": "merged.pdf" }
}

Split one PDF into three items (pages 1–3, page 7, page 9 to end):

{
  "resource": "document",
  "operation": "split",
  "binaryPropertyName": "data",
  "pageRanges": "1-3,7,9-"
}

Generate a PDF from a JSON template (no input binary needed):

{
  "resource": "generate",
  "operation": "fromTemplate",
  "template": "{\n  \"content\": [\n    { \"type\": \"heading\", \"level\": 1, \"text\": \"Invoice #1042\" },\n    { \"type\": \"paragraph\", \"text\": \"Thank you for your business.\" },\n    { \"type\": \"table\", \"headers\": [\"Item\", \"Qty\", \"Price\"], \"rows\": [[\"Widget\", \"2\", \"$10.00\"]] }\n  ],\n  \"pageNumbers\": true\n}",
  "options": { "outputFileName": "invoice.pdf" }
}

Limits

• Input size: binaries over 100 MB are rejected with a clear error rather
  than risking an out-of-memory crash.
• Roadmap: text extraction and PDF encryption are planned but need
  engines (pdfjs-dist, qpdf/WASM) that are currently incompatible with
  n8n's community-node security rules; they'll return if a scanner-compatible
  engine becomes available.
• Heavy documents: operations are CPU-bound and run in-process. Very
  large documents processed concurrently with other busy workloads on the
  same instance may add latency to that instance.
• Form → Fill Form and emoji: pdf-lib renders one form field's whole
  appearance with a single font, unlike Generate/Stamp's per-character emoji
  fallback. A field value containing an emoji or other pictographic
  character throws a clear error naming the field, rather than silently
  producing a blank appearance.

Credentials

None required. All operations run locally on binary data already in your
workflow.

Compatibility

Requires n8n with community-node support (n8nNodesApiVersion: 1). No
minimum n8n version is pinned; the node relies only on standard binary-data
helpers.

Resources

• n8n community nodes documentation
• Report issues

Version history

• 0.2.3 – 0.2.7: Documentation and metadata updates — README refresh with
  workflow screenshot and version history, author contact metadata, and the
  codex node identifier fixed to the fully-qualified format
  (n8n-nodes-pdf.pdfToolkit). No functional changes. (0.2.5 was skipped.)
• 0.2.2: Removed Extract → Text and the Secure resource (Encrypt,
  Decrypt, Set Permissions) from the node UI; they were never functional,
  engine-blocked stubs since 0.2.0. Five resources / 18 operations, all
  functional.
• 0.2.1: Unicode + emoji text support. Generate, Text Watermark, Page
  Numbers, and Fill Form now draw text with bundled Noto Sans/Noto Sans
  Mono/Noto Emoji fonts (Latin Extended, Cyrillic, Greek, monochrome emoji)
  instead of the WinAnsi-only built-in fonts.
• 0.2.0: 18 of 22 operations implemented (all of Document, Generate,
  Form, and Stamp; Extract Metadata, Embedded Images, and Page Count),
  covered by an automated test suite. Zero runtime dependencies.
• 0.1.0: Initial scaffold. Node UI and routing for 6 resources / 22
  operations.