Everything happens on your device.
Nothing goes outside.

Q: Something broke. Who do I tell?

Email hello@pdfmcp.org — a human reads it. Service health is public on the status page.

74 PDF tools — merge, convert, compress, OCR, sign, redact — that run in your browser, not on our servers. Your files are never uploaded. And for agents and developers, the same tools are a free MCP server and API.

🔒 Private by design — files never leave your device 🆓 100% free — no sign-up, no watermark 🤖 250 free API ops/day + burst on spare capacity 🗑️ Server files deleted within 1 hour — content never logged

Browse the tools Connect your AI agent →

Most-loved tools

PDF to Word (DOCX) API · MCPConvert a PDF to an editable Word document. Text-based conversion: paragraphs and headings are reconstructed from the text layer; exact layout, tables and images are not preserved.Merge PDFs On-deviceCombine multiple PDF files into one, in upload order.Compress PDF On-deviceReduce file size with Ghostscript. Presets: light (printer quality), medium (ebook), heavy (screen, smallest).Split PDF On-deviceSplit a PDF by page ranges (ZIP of PDFs) or one PDF per page.PDF to text On-deviceExtract plain text, one form-feed separated section per page.Images to PDF On-deviceCombine images (JPG/PNG/WebP/TIFF/GIF) into a single PDF.OCR — searchable PDF API · MCPMake a scanned PDF searchable: pages without a text layer are OCR'd with Tesseract and gain an invisible text layer; pages that already have text are passed through and reported as skipped.Strip AI metadata & provenance On-deviceRemove AI-generation provenance and identifying metadata from a PDF or image (JPEG/PNG/WebP): C2PA Content Credentials, XMP/EXIF/IPTC AI markers (incl. IPTC DigitalSourceType trainedAlgorithmicMedia), generator Software/Creator tags, EXIF GPS and device data. Metadata only — content and pixels are never modified, image pixels are not re-encoded, and pixel-embedded invisible watermarks (e.g. SynthID) are NOT removed: they live in the image data itself, not in metadata. The result reports exactly what was found and removed, by category.

Convert PDF to Word is our most-used tool — followed closely by merging PDFs, compressing PDFs and making scans searchable with OCR.

Organize PDF

Merge PDFs On-deviceCombine multiple PDF files into one, in upload order.Split PDF On-deviceSplit a PDF by page ranges (ZIP of PDFs) or one PDF per page.Rotate pages On-deviceRotate pages by 90, 180 or 270 degrees.Delete pages On-deviceRemove the selected pages from a PDF.Extract pages On-deviceCreate a new PDF containing only the selected pages.Reorder pages On-deviceRearrange pages into the given 1-based order (a permutation of all pages).

See all organize pdf tools →

Convert to PDF

Images to PDF On-deviceCombine images (JPG/PNG/WebP/TIFF/GIF) into a single PDF.HTML to PDF API · MCPRender an HTML document to PDF. Fully offline: external resources (images, stylesheets, scripts) are never fetched — inline everything you need.Markdown to PDF API · MCPRender a Markdown document (GitHub-flavoured) to a typeset PDF.Text to PDF On-deviceRender a plain-text file to a PDF (monospaced, line-wrapped).Office document to PDF API · MCPConvert Word/Excel/PowerPoint/OpenDocument files (docx, xlsx, pptx, odt, ods, odp, doc, xls, ppt, rtf) to PDF via the LibreOffice backend.EPUB to PDF API · MCPConvert an EPUB ebook to a typeset PDF (text-focused: chapter text and headings in spine order; embedded images are not rendered).

See all convert to pdf tools →

Convert from PDF

PDF to Word (DOCX) API · MCPConvert a PDF to an editable Word document. Text-based conversion: paragraphs and headings are reconstructed from the text layer; exact layout, tables and images are not preserved.PDF to text On-deviceExtract plain text, one form-feed separated section per page.PDF to images On-deviceRender each page to an image (ZIP).PDF to Markdown On-deviceExtract text as Markdown, with headings inferred from font sizes.Extract text as JSON On-devicePer-page structured text: blocks with bounding boxes, ready for agents/LLMs.PDF to HTML On-deviceConvert a PDF to a single HTML document, one section per page, preserving positioned text.

See all convert from pdf tools →

Optimize PDF

Compress PDF On-deviceReduce file size with Ghostscript. Presets: light (printer quality), medium (ebook), heavy (screen, smallest).Repair PDF API · MCPRewrite and repair a damaged PDF with qpdf.Flatten PDF On-deviceBake form fields and annotations into the page content.Convert to PDF/A API · MCPConvert a PDF to the PDF/A archival standard with Ghostscript.Linearize PDF API · MCPRewrite a PDF linearized ('Fast Web View') so first pages display before the whole file downloads.Optimize images in PDF API · MCPDownscale and recompress oversized embedded images, and drop duplicate image data.

See all optimize pdf tools →

Protect & sign

Encrypt PDF API · MCPProtect a PDF with a user password (AES-256).Unlock PDF API · MCPRemove password protection (the current password is required).Strip AI metadata & provenance On-deviceRemove AI-generation provenance and identifying metadata from a PDF or image (JPEG/PNG/WebP): C2PA Content Credentials, XMP/EXIF/IPTC AI markers (incl. IPTC DigitalSourceType trainedAlgorithmicMedia), generator Software/Creator tags, EXIF GPS and device data. Metadata only — content and pixels are never modified, image pixels are not re-encoded, and pixel-embedded invisible watermarks (e.g. SynthID) are NOT removed: they live in the image data itself, not in metadata. The result reports exactly what was found and removed, by category.Sanitize PDF On-deviceStrip metadata (incl. XMP), JavaScript, embedded files/attachments, document actions and link/file actions — the safe-to-share version of a PDF.Sign PDF (visual signature) On-devicePlace a visual signature on a PDF: upload the signature image as the SECOND file, or let the signer name be typeset. A self-signed certificate (SHA-256 fingerprint) is generated and recorded in the document metadata. This is an electronic signature stamp + certificate record — not a qualified/PKI document signature.Set PDF permissions API · MCPEncrypt a PDF with an owner password and fine-grained permissions (print, copy, modify, annotate, ...).

See all protect & sign tools →

Edit & mark up

Watermark PDF On-deviceStamp diagonal text across every page.Add page numbers On-deviceStamp page numbers, format 'Page {n} of {total}'.Add header & footer On-deviceStamp a header and/or footer line on every page. {page} and {total} placeholders are substituted per page.Bates numbering On-deviceStamp sequential Bates numbers (prefix + zero-padded counter) on every page.Fill PDF form On-deviceFill AcroForm fields by name. Values: strings for text fields, true/false for checkboxes, numbers are stringified.Read form fields On-deviceList all AcroForm fields with names, types, current values and locations (JSON).

See all edit & mark up tools →

Extract & OCR

OCR — searchable PDF API · MCPMake a scanned PDF searchable: pages without a text layer are OCR'd with Tesseract and gain an invisible text layer; pages that already have text are passed through and reported as skipped.PDF info & metadata On-devicePage count, page sizes, metadata, encryption status as JSON.Extract images API · MCPPull all embedded images out of a PDF (ZIP, original formats).Extract tables (JSON + CSV) API · MCPDetect tables and return them as tables.json plus one CSV per table (ZIP).Extract bookmarks / TOC On-deviceReturn the document outline as JSON: level, title, target page.Extract attachments On-devicePull all embedded file attachments out of the PDF (ZIP).

See all extract & ocr tools →

When a job does need a server, the rules don't change

Some tools need server-grade machinery — OCR at scale, Office conversion, cryptography. Those run on our MCP/API servers, under the second half of the promise: we don't keep your stuff. Files are virus-scanned before processing, stored encrypted, and deleted within an hour. Content is never logged. And it's free and generous: 250 operations a day guaranteed, paced at 60 an hour so everyone's share stays available, plus unlimited burst whenever the machine is idle.

Read the full free policy · Read the privacy commitments · See live service status

Give your AI agent hands: the free MCP server

Claude, ChatGPT, Cursor, VS Code, Windsurf, Codex, Gemini CLI and more can call every pdfmcp operation natively through MCP. One free key, two minutes of setup:

{
  "mcpServers": {
    "pdfmcp": {
      "url": "https://mcp.pdfmcp.org/mcp",
      "headers": {
        "X-API-Key": "pdfmcp_live_YOUR_KEY"
      }
    }
  }
}

Setup guides for 13 MCP clients · Get your free key · Prefer plain REST? Start here

Questions

Is pdfmcp really free?

Yes. The web tools are free with no sign-up and no watermark. API/MCP keys get 250 guaranteed operations a day plus unlimited burst on spare capacity. There's no paid tier at launch — a donation box (currently disabled) is the whole business model for now.

Where do my files go?

On web tools: nowhere. Everything happens on your device — the file never leaves your computer. On the API/MCP servers: into an encrypted temp volume, and they're auto-deleted within 1 hour. Content is never logged.

What is the MCP server?

MCP (Model Context Protocol) lets AI agents — Claude, ChatGPT, Cursor, VS Code and others — call tools directly. Our MCP server exposes all 74 operations, so your agent can merge, OCR, convert and redact PDFs itself. Free, with the same key as the REST API.

Why is it free? What's the catch?

The web tools cost us nothing (your device does the work). The server is one modest machine run under a generous fair-use policy. No ads, no trackers, no selling data — we couldn't sell your data anyway, because on web tools we never receive it.

Something broke. Who do I tell?

Email [email protected] — a human reads it. Service health is public on the status page.

Everything happens on your device.Nothing goes outside.