The short version
- You select a document in your browser.
- The document and image tools generate visual previews locally, without uploading the selected files.
- Merge, Split, Organize, supported image-to-PDF exports, and supported PDF-to-image exports run on your device. If local processing cannot finish, files automatically use temporary server processing. The processing dialog shows the switch and lets you cancel.
- When you start a server operation, the browser sends the document over HTTPS to the PDFMK API.
- The API uses a short-lived temporary working directory to process the request and prepare the result.
- Completed results may be saved to private Cloudflare R2 storage. The API returns a signed link valid for three days and removes the local task files after upload. Legacy file responses are removed after sending; errors also trigger cleanup.
Uploaded document content is not used to create advertising profiles, train models, or build a permanent document library. PDFMK does not currently provide user accounts or permanent cloud document storage. Results in R2 are scheduled for deletion after three days; physical deletion typically follows within 24 hours. Keep signed links private.
Processing by tool
| Tool | Processing behavior | Result |
|---|
| Merge | Common PDFs are merged locally. PDFs with document-level features or large memory requirements automatically use the existing server merger when local processing cannot finish. | One PDF |
| Split | Selected pages and ranges are extracted locally into a PDF or ZIP. If local processing cannot finish, files automatically use temporary server processing. | PDF or ZIP |
| Organize | Page order, rotation, omission, and duplication are applied locally. If local processing cannot finish, files automatically use the existing server processor. | One PDF |
| Watermark | Only watermark text and appearance settings are sent to prepare a small asset. Your PDF is processed locally; compatibility failures automatically use temporary full-document server processing. | One PDF |
| PDF to Word | The first-page preview and text-layer check stay local. After you click Convert to Word, the selected PDF pages are processed in a temporary working directory and cleaned up after the result is uploaded or the direct response finishes. | One DOCX |
| PDF to images | Supported PDFs and ZIP inputs render locally using an engine downloaded on first conversion and cached by the browser. Engine, compatibility, or resource failures automatically use temporary server processing. | ZIP of PNG files |
| Images to PDF | Common images are placed on PDF pages locally, using the chosen layout and background. Unsupported decoders, CMYK JPEGs, animation, and resource limits automatically fall back to temporary server processing. | One PDF |
What you should not upload
No online processor can promise zero risk. Do not upload a document when law, contract, company policy, or professional duty requires it to stay on a managed device or an approved internal system. That can include medical records, identity documents, privileged legal material, export-controlled data, and unredacted financial records. Use an approved offline tool for those files.
Operational safeguards and limits
- HTTPS protects documents while they travel between a supported browser and the service.
- Temporary-path cleanup runs after successful and failed processing paths.
- ZIP inputs are inspected before extraction and non-PDF entries are ignored by the image converter.
- Password-protected or damaged PDFs may be rejected rather than partially processed.
- Practical size limits depend on page count, image complexity, selected DPI, and available server resources.
Read the Privacy Policy for information about technical logs, advertising, and data rights. Report a suspected vulnerability through the contact page.