PDF Tools Skill Review: Worth Installing in 2026?

Anthropic shipped the PDF Tools skill for Claude Code about six months ago. It's gotten featured in most "must-install skills" lists, but those lists rarely tell you where the tool actually breaks down. I've been using it daily for legal contract review, academic paper extraction, and converting product manuals into searchable notes. Here's what's worth knowing before you install.

What it does, in plain terms

PDF Tools is a Claude Code skill that lets you ask Claude to do anything you'd otherwise do with Adobe Acrobat Pro: read text, extract tables, run OCR on scanned pages, merge multiple files, split one file into chapters, rotate pages, add watermarks, fill forms.

You ask in plain English. Claude figures out which underlying Python library to call (pdfplumber, PyMuPDF, Tesseract for OCR), runs it on your local file, and gives you the result. No upload to a cloud service. No subscription beyond what you already pay for Claude.

Install: /plugin install pdf-tools GitHub: github.com/anthropics/claude-skills Cost: Free. Requires Claude Code (Claude Pro at $20/month or higher).

What works well

Text extraction on native PDFs

PDFs that were exported from Word, LaTeX, or any modern publishing tool are handled flawlessly. Page 1 to page 800, equal quality throughout. I ran it on a 412-page open-source software documentation PDF and the extracted text was a clean Markdown file with headings preserved.

Speed: about 90 seconds for the 412-page extraction on an M2 MacBook Air. Faster than Adobe's "Export to Word" on the same machine.

Table extraction

This is where most PDF tools fail badly. PDF Tools, using pdfplumber under the hood, handles tables better than 80% of competitors I've tested. A 47-row pricing table from a SaaS comparison PDF came out as a clean CSV on the first attempt. Multi-page tables (a table that spans pages 4-7) got correctly stitched into one continuous output, the part most tools mess up.

Where it still struggles: tables with merged cells or visually-complex headers. You'll occasionally get rows with shifted columns. Always sanity-check table outputs before pasting into a spreadsheet you care about.

Merge and split

Boring, but reliable. Merging five PDFs into one preserves bookmarks and metadata. Splitting a 200-page contract into individual sections by chapter detection works, though you sometimes need to specify the page ranges manually if the headings aren't picked up.

What's mediocre

OCR on scanned PDFs

This is the marketed killer feature and it's where reality diverges from the hype. PDF Tools uses Tesseract for OCR, which is open source and decent but noticeably behind commercial OCR engines (Adobe's, ABBYY FineReader).

Test case: a 30-page scanned textbook chapter at decent quality (300 DPI, modern font). PDF Tools extracted text with about 94% character accuracy. Adobe Acrobat Pro on the same file: 99%+. The difference matters, that 5% gap on a 30-page document means about 50 misread words, and they cluster in places that change meaning (numbers, technical terms, citations).

For older scans, faded photocopies, or anything below 200 DPI, accuracy drops fast. I tested a 1980s academic paper scan and got around 80% accuracy, usable for getting the gist, useless for citation work.

Verdict: Fine for modern, well-scanned documents. For archival or low-quality scans, you'll want a dedicated commercial OCR tool.

Form filling

The skill claims to fill PDF forms, but interactive PDF forms (the ones with proper form fields) work better than scanned "forms" that are actually just images with lines on them. For the latter, the skill tries to OCR the field locations and place text on top, with mixed results.

For most form-filling use cases I'd recommend pdf.gov-grade tools or Adobe Acrobat. PDF Tools' form support feels like a v0.5 feature.

What breaks

Encrypted/protected PDFs

If a PDF has password protection or DRM, PDF Tools can't open it. The skill will tell you this, which is honest, but it can't crack protections. You'll need to remove protection upstream (with permission from the document owner) before using the skill.

Extremely large files

I tested a 1.2GB PDF (a corporate annual report with hundreds of embedded charts). PDF Tools handled it but processing took 14 minutes for full text extraction. Memory usage spiked to 6GB. On laptops with 8GB of RAM this caused thrashing. If you regularly work with huge PDFs, do it on a desktop or split files first.

Multi-column academic papers

Two-column journal article layouts (standard for IEEE, ACM publications) sometimes produce reading-order errors, the extractor reads across columns rather than down each column. You can work around this by specifying column-aware extraction, but the default behavior is wrong about 30% of the time on dense two-column PDFs.

How it compares to alternatives

The pattern: PDF Tools wins on workflow speed and table extraction, loses on OCR quality and form handling.

Who should install it

Install if:

• You handle 20+ PDFs a week in research, legal, or content work
• Most of your PDFs are modern, native (not scanned) documents
• You already pay for Claude Pro and use Claude Code for other work
• You want batch operations via plain English rather than scripting
• You care about local processing (no cloud uploads)

Skip if:

• You mostly work with old scans, faded photocopies, or low-DPI archival material, pay for Adobe Acrobat instead
• Your PDF work is occasional (a few documents per month) — the free online tools cover this
• You don't already use Claude Code for anything else, the install overhead isn't worth one skill

Real-world workflow examples

Legal contract review

I dropped a 60-page NDA into a folder and asked Claude: "extract every clause that mentions liability, indemnification, or termination, and show me the page numbers." Got back a 2-page summary in 30 seconds with accurate page references. Replaced what used to be 45 minutes of skimming.

Academic literature review

For a research project I needed to extract the methodology section from 23 papers. With PDF Tools: "from every PDF in this folder, extract just the Methodology section and save them all as one Markdown file." Done in under 4 minutes. The manual version would have taken an afternoon.

Operations documentation

A small business owner I work with had 14 supplier contracts as PDFs and wanted a comparison table of payment terms. PDF Tools generated the comparison CSV in about 90 seconds. Two columns needed manual cleanup (one supplier used unusual formatting for the payment terms). Still saved roughly 3 hours of manual work.

Installation walkthrough

If you decide to install:

1. Install Claude Code from claude.ai/code if you don't have it 2. Open Claude Code in a terminal (or via your IDE's Claude Code integration) 3. Run /plugin install pdf-tools 4. First-time use will prompt you to install Python dependencies (pdfplumber, PyMuPDF, Tesseract). On Mac and Linux this is automatic. On Windows, Tesseract needs a manual installer download from UB Mannheim. 5. Verify by asking Claude: "Read the PDF at ./test.pdf and tell me how many pages it has."

Total setup time: 5-10 minutes depending on your OS.

Security notes

• All processing happens locally. PDFs you analyze are not sent to Anthropic or any third party.
• The skill uses standard open-source libraries (pdfplumber, PyMuPDF, Tesseract). These are widely used and security-audited.
• If you process confidential documents, Claude Code itself sees the extracted text as part of its context. This is the same trust model as using Claude for any work, review Anthropic's data policy if your documents are sensitive.

FAQ

Q: Does it work offline? Partly. The PDF processing happens locally, but Claude itself runs through Anthropic's API. You need an internet connection for the conversational part, even though the PDF never leaves your machine.

Q: What about handwritten PDFs? Tesseract handles printed text well, handwritten text badly. For handwriting recognition you'd want a specialized tool. Microsoft Lens or Google Lens are the consumer-grade options.

Q: Can I use it with Claude on the website (not Claude Code)? No. This skill specifically requires Claude Code. Claude.ai handles PDF uploads natively but doesn't have the batch/automation capabilities the skill enables.

Q: How does it compare to Anthropic's own PDF support in Claude.ai? Claude.ai's built-in PDF reading is great for single-document Q&A but caps out at smaller file sizes and doesn't do batch operations, table extraction to CSV, or file conversions. PDF Tools is the upgrade if you need any of those.

Q: Will my professor / boss / lawyer accept work done with this? The tool produces standard PDF/Word/CSV/Markdown outputs that look identical to manually-produced files. Whether AI assistance is disclosed is a separate ethical question depending on your context.

Verdict

PDF Tools is the right install for anyone whose work touches PDFs daily, researchers, lawyers, accountants, students with heavy reading loads, content marketers extracting source material. It's well-built where it matters most (text extraction, table extraction, batch operations) and clearly weaker in two specific places (OCR on poor-quality scans, complex form handling) where you can route around it with a single dedicated tool.

For the 80% of PDF tasks most knowledge workers run into, it's faster and more pleasant than Adobe Acrobat, and it's free if you already use Claude. That's enough to recommend it.

Rating: 8/10. Best-in-class for native PDF workflows, average for OCR-heavy work.