Scan the stack once.
Put the whole pile through your scanner. Upload the resulting PDF once. Papmil separates the individual documents, names each file, and hands them back ready for ordinary folders or Paperless-ngx.
Two free workflow credits, valid for two weeks · PDF up to 15 MB · Files deleted after 30 days
Each split result keeps travelling through the chain.
Real production totals. Small numbers are welcome: this is a working proof of concept, not manufactured social proof.
The first useful job
The current version is built for the pile of letters, invoices, statements, and notices that you would otherwise separate and rename by hand.
Feed a large pile of paper through the scanner as one PDF. No careful batching between documents.
The workflow makes the scan searchable, finds document boundaries, and creates a useful filename for each result.
Download the separate PDFs and put them in normal folders, or send them into a proper archive such as Paperless-ngx.
The live workflow
Every operation hands its output to the next. If a step splits one PDF into five documents, all five continue through the remaining steps.
Papmil runs one ready-made four-step workflow. The engine supports longer and different pipelines; a self-service builder is not available yet.
Reduce the scan's file size before downstream processing.
Run OCR to create searchable PDFs. Language selection is not wired yet.
Find likely document boundaries so a mixed scan becomes separate PDFs.
Use date, sender, document type, and subject to create practical filenames.
No asterisk
Your PDF is stored temporarily in AWS S3. Papmil's worker downloads it, runs the PDF tools, and uses an Ollama model on that processing machine for splitting and naming.
Document content is not sent to OpenAI, Anthropic, Google Gemini, or another external AI API. Source, intermediate, and result files are automatically deleted 30 days after the workflow finishes.
The experiment is live
