Every bank statement, warranty slip, and tax document you scan into Paperless-ngx becomes instantly searchable, because the software runs OCR on each one and full-text indexes the result. Type a supplier name or a number and the matching PDF surfaces in seconds, all on a server you own, with nothing handed to a cloud service. Self-hosting it is a Docker job that a modest home machine handles comfortably. Here is the full setup.
Quick Answer
Paperless-ngx digitises documents by scanning them, running Tesseract OCR to extract the text, and full-text indexing everything so your archive becomes searchable. You self-host it with Docker Compose, which spins up the web app, a PostgreSQL database, and Redis in one command. The full stack idles at roughly 600 to 900 MB of RAM, so a 2 GB machine, even a Raspberry Pi 4, runs it fine.
What Paperless-ngx does for you
The pitch is simple, paper and loose PDFs become a tidy, searchable library. You feed documents in by scanning or dropping files into a watched folder, Paperless-ngx OCRs them with Tesseract so the text inside becomes machine-readable, then tags, dates and indexes them. From there you search by content, filter by correspondent, or browse by tag. Tesseract supports over 100 languages, and you can specify several at once for mixed-language households.
The whole point of self-hosting is control. Your financial and personal documents stay on your own hardware, not on someone else's servers, and there is no subscription.
What you need to run it
Paperless-ngx is undemanding. It supports amd64, arm and arm64 hardware, so it runs on everything from a Raspberry Pi 4 to a mini PC to a repurposed desktop. The full stack uses around 600 to 900 MB of RAM at idle, so 2 GB is enough to start, though more memory and a faster disk speed up OCR on big batches.
A small, low-power machine that can stay on around the clock is the ideal host, you want it always available so documents process the moment you add them. Storage is the part to size generously, since this becomes your permanent archive.
Step-by-step Docker setup
- Install Docker and Docker Compose on your host machine. This is the only prerequisite, Paperless-ngx ships entirely as containers.
- Pull the official Paperless-ngx compose configuration. The standard deployment defines three services, the web application, a PostgreSQL database for metadata, and a Redis instance for the task queue.
- Edit the environment settings. Set your OCR language, a strong admin password, and the storage paths where documents and the consume folder will live.
- Bring the stack up with
docker compose up -d. Docker downloads the images and starts the containers in the background. - Open the web interface in a browser, by default on port 8000, and log in with the admin account you configured.
Two optional services, Gotenberg and Tika, improve conversion of office files and emails. Add them if you plan to feed in more than scans and PDFs.
Getting documents in and finding them again
Once it is running, the workflow is the easy part. Drop files into the consume folder or upload through the web interface, and Paperless-ngx OCRs and files them automatically. Set up correspondents, document types and tags, and you can let rules apply them on ingest so new documents land pre-sorted.
Searching is where the effort pays off, because the full-text index means you can find a document by a phrase buried inside it, not just by its filename. On the 2026 stack you can even wire in a local language model for automatic classification and smarter tagging, keeping that intelligence on your own hardware too.
For the host itself, a quiet always-on machine is the natural fit -- something small, efficient, and happy running around the clock. The mini PCs at Evetech are sized precisely for this role, drawing little power and fitting anywhere. If you have a more capable repurposed desktop in mind for double duty, the popular PC builds on Evetech show what local home-server builders are choosing right now.
Frequently Asked Questions
What hardware do I need to run Paperless-ngx?
Very little. It runs on amd64, arm and arm64, including a Raspberry Pi 4, and the full stack idles at roughly 600 to 900 MB of RAM. A 2 GB machine handles it, though more memory and a faster disk speed up OCR on large batches.
Does Paperless-ngx need an internet connection?
No, it runs entirely on your own network, which is the main appeal. You only need internet to install or update the containers, the OCR, indexing and search all happen locally on your server.
How does the OCR work?
Paperless-ngx uses Tesseract, the leading open-source OCR engine, to read the text inside your scans and PDFs. That extracted text is then full-text indexed so you can search documents by their contents, not just filenames.
Can I run it on a Raspberry Pi?
Yes. Paperless-ngx supports arm64 and runs on a Raspberry Pi 4, though OCR is slower than on a more powerful mini PC. For a steady trickle of household documents, a Pi is perfectly capable.
Is my data safe self-hosting this?
Your documents stay on hardware you control, which is more private than a cloud service, but you become responsible for backups. Back up the document store and the database regularly, since this is your permanent archive.
Turn your paper pile into a searchable archive you own. Pick a quiet, always-on host from the mini PC range at Evetech and get Paperless-ngx running this weekend.