Self-hosted document management
Thaw vs paperless-ngx
Technical households who ran paperless-ngx in Docker on a home server or a Mac mini, loved the consumption folder, and got tired of being the sysadmin.
paperless-ngx is excellent software and the right answer if you want to own the box. It OCRs, tags, and files with rules you write. What it does not do is understand the document: it will not read the term end off a policy, link a vet record to the dog, or notice that 2021's return is missing. And it needs a computer that is always on. Thaw imports its export ZIP directly. We built Thaw after running paperless-ngx ourselves.
Side by side
| Thaw | paperless-ngx | |
|---|---|---|
| Hosting | Hosted. Nothing to run at home. | You host it: Docker, a database, a Redis, backups, updates. The Mac mini stays on. |
| How paper gets in | Email-in address per family or business, scanner-to-email, browser upload, folder import, Evernote and paperless-ngx imports. Dropbox and Drive folder watch next. | Consumption folder, email fetch from a mailbox, the web upload, mobile apps from the community. |
| Classification | Three axes (category, type, form) read from the page by a model, with a registry that grows from what it sees. Never "uncategorized". | Correspondents, document types and tags you define, matched by rules or auto-matching on OCR text. Since 3.x an opt-in AI can suggest them and chat about a document, with your own OpenAI key or a local Ollama. |
| Facts and dates | Every document is read once: category, type, title, facts with roles, the people & things it concerns, every date. Uncertain reads are flagged, never queued. Every renewal, expiry, due date and notice period becomes an item with what happens if missed and what to do, from the document itself. Resolved by the next document. | Created date and custom fields you fill in, with AI suggestions if you turn them on. No obligations. |
| People & things | Every document linked to who and what it concerns; a page per person, vehicle, property. | Correspondents (senders). No notion of the subject. |
| What is missing | Knows what a person, vehicle, property, pet or business should have on file; lists what is missing, lapsed, expired or duplicated, worst first, with where to get it. | None. |
| Data ownership | Originals in a folder tree named the way Thaw names them, plus facts as CSV and JSON. Continuous mirror to your own cloud folder on the roadmap. Your keys are yours to destroy. | Everything on your disk, in the open. Total control. |
| Cost | Free tier, then from $8 a month. | Free software, plus the hardware, the electricity and your Saturday. |
Stay with paperless-ngx if
- You want every byte on hardware you own and no third party in the loop.
- You enjoy running it.
- Your workflow is entirely rule-based and it already works.
Move to Thaw if
- You want the machine off.
- You want the documents understood, not just OCRed and tagged.
- You want someone in the household who is not you to be able to find things and be told what is due.
Moving over
Run the document exporter (document_exporter) to a folder and zip it. Upload the ZIP to your Thaw import bucket or drop it in the browser if it fits. Thaw reads manifest.json for correspondents, document types, tags and created dates and passes them to the reader as hints. The originals are imported, not the archived OCR copies.
Point it at the pile and walk away.
Early access is open to a small group of households and owner-run businesses. Tell us a little about your paper and we will send an invitation.