Paperless-ngx
FreeScan the paper in once, then search for it instead of digging through drawers
A self-hosted document management system. Scan or photograph paper documents and it runs OCR, infers document type and date, files them by rule-based tagging, and makes every word searchable afterwards. Contracts, invoices, insurance policies, manuals and certificates all work this way.
Why we picked it
Every household has a drawer full of paper: contracts, insurance policies, invoices, repair records, manuals, copies of documents. When you need one, you know it is "in there somewhere" — and finding it takes twenty minutes.
Paperless-ngx solves that. 44,255 stars, GPL-3.0, the most mature open-source answer in the field.
The workflow is simple:
- Put it in. A scanner, a phone photo, or an electronic PDF dropped into a watched folder.
- It processes automatically. OCR extracts the text across many languages, the date is inferred from the content, the issuing organization is recognized, and your rules apply tags and categories.
- Search for it later. Type a phrase from the contract and it searches the full text of every document, returning results instantly.
Key capabilities:
- OCR makes scans searchable. That is the core value — a scanned image contains no findable text until OCR runs. The output is a PDF with a text layer, so any reader lets you select the words.
- Automatic classification rules such as "title contains 'electricity' → tag as Utilities, correspondent is the power company", so similar documents file themselves from then on.
- Originals are preserved. It does not modify your source files; OCR results are additive.
- Full-text search with fuzzy matching and advanced queries.
- A mobile-friendly web interface for uploading straight from a phone camera.
To be clear about:
- Deployment is required and not lightweight. Docker Compose is the official path, with a database, Redis, an OCR engine and the web service. The documentation is thorough, but some self-hosting experience helps.
- OCR is CPU-hungry. Importing hundreds of documents runs for a long time; a Raspberry Pi manages but slowly, so give it a capable machine if you can.
- Non-English OCR needs the right language packs, which may not be installed by default. Specifying the language in the configuration improves accuracy considerably.
- It manages documents, not files. Ideal for contracts and invoices that need long-term archival retrieval, wrong as a general cloud drive — that is Nextcloud's job.
- Backups are critical. Once you discard the paper originals, this system holds the only copy. Configure your backup strategy before you start scanning at scale — at minimum one offsite or offline copy.
Who it suits: anyone digitizing household paperwork, freelancers and small businesses retaining business records, and everyone tired of knowing they have a document but being unable to find it.