Paperless-ngx: a self-hosted home for your paperwork

Paperless-ngx is a community-supported, open-source document management system that turns physical documents into a searchable online archive. The code is GPL-3.0 licensed.

Here is what happens when a file arrives. By default, Paperless runs OCR only if the document has no text, creates a PDF/A copy of scanned or image-based documents for long-term storage, and matches tags, correspondents, and document types from the text it read. It keeps the original file untouched next to the archived version. The project says the matching uses machine learning, and its OCR engine is Tesseract, which handles more than 100 languages.

You then find things with full-text search: it suggests words from your own documents, ranks results by relevance, marks the matching parts, and can show similar documents. There is also email import, multi-user permissions, and shareable links with optional expiry. AI features that use large language models are opt-in and off by default, and so is remote OCR through Azure AI. I would leave both off for financial paperwork.

Your bitcoin leaves a paper trail: exchange statements, purchase records and cost-basis exports for taxes, hardware wallet receipts, shipping confirmations with your home address on them, and the letter of instruction your heirs will need. Together they map what you own and where you live. A cloud drive or a scanning app's cloud puts that map behind one account password, on a company's servers.

Paperless-ngx keeps it on a machine in your house. The project says your data is stored locally on your server and never transmitted or shared unless you choose to share it. My inheritance guide explains what your heirs need to find, and Paperless is a good place to keep the letter and the non-secret instructions. If you want plain file sync instead, Nextcloud fills that role.

One hard rule: never scan or store a seed phrase or passphrase in Paperless, in a photo, or anywhere else digital. A scan of your words turns a private backup into a file that sits on a disk, in backups, and possibly in a phone photo roll. Keep the words on paper or metal, as the seed phrase backup guide explains. Paperless can hold the letter that tells your heirs where the words are, but not the words.

The project says Docker is the best option for most people: faster to set up and easier to maintain than bare metal, which runs on Linux only. Its FAQ says Docker images are available for arm64 and that it runs on a Raspberry Pi, but parts such as OCR run very slow there, so give it a mini PC rather than the smallest board you own. My home server guide covers hardware. The setup docs give the current Docker Compose files and steps.

On umbrelOS, Paperless-ngx is listed in the Umbrel App Store. On StartOS it is a community package in Start9's community registry rather than an official Start9 one, and as of November 2026 it trails the upstream release. Everywhere else, Docker Compose is the route.

The main route is the consumption directory. Paperless watches that folder, runs each new file through OCR and matching, stores the result, and then removes the file from the folder, so treat it as a temporary inbox. You can also drag files into the web interface or let it pull attachments from email accounts.

To make a network scanner drop files into that folder, the docs say you need a service on the server to accept them, typically an FTP server such as ProFTPD or a Windows-style folder share through Samba. If the folder sits on NFS, turn on polling, because that file system does not send the change notifications Paperless relies on.

For phones, the project does not ship its own app. It points to a user-maintained list of related projects and says the team does not endorse or support them. The list includes Swift Paperless for iOS and Paperless Mobile for Android. These are third-party apps that hold a login to your document server, so check each one's code and privacy practices before pointing it at financial records. The project also keeps a scanner and software recommendations page.

Paperless-ngx does not encrypt your documents on disk. Its README warns that information is stored in clear text without encryption, that it should never run on an untrusted host, and that the safest way to run it is on a local server in your own home with backups in place. The features list agrees that documents are stored plain on disk.

That makes the disk the thing to protect. Turn on full-disk encryption on the server so a stolen or recycled drive is unreadable, keep the machine at home, and keep the web interface off the internet. Full-disk encryption does not help while the server is running and unlocked, so a compromised server still exposes everything. Encrypt your backups separately, and keep a strong password on every account.

Your scans may be the only digital copy of a record, so back them up. The project's backup docs name the document exporter as the option that works for any install. It writes your documents, thumbnails, metadata, and database contents to a folder, and it can update an existing export, so incremental backups with rsync work. The docs give the current commands for running it and for restoring.