The self-hosted Paperless alternative
that's 182× lighter.
Scan it, drop it, find it in a millisecond. An open-source document archive with OCR and instant full-text search, in one container that idles at 5 MB of RAM.
docker run -d -p 8000:8000 -v papiro-data:/data -e PAPIRO_ADMIN_USER=admin -e PAPIRO_ADMIN_PASSWORD=change-me ghcr.io/elpablazo/papiro
Paperless-ngx boots Django five times over and keeps Redis, PostgreSQL, Tika and Gotenberg running next to it. Papiro is one Rust process with SQLite inside. Measured on the same machine with the same documents.
Every number comes from a script you can run. Method and raw data, including where Paperless wins.
Built for the person who keeps the household paperwork, not for an IT department.
Results as you type, with the matching words highlighted. Filter by folder, tag, correspondent, type and date.
Suggests folder, type and tags from documents you already filed, and spots expiry dates and totals. No AI service, nothing leaves your server.
An installable app: snap a receipt with the camera, or (on Android) share a PDF from any app straight into your archive.
A "due dates" view for passports, insurance and leases, plus totals over any list of receipts.
Crash-safe ingest, a daily SHA-256 check of every document, and backups with 14 days of history.
REST API with OpenAPI, signed webhooks, and an MCP server so Claude can search and file for you.
Bring everything over in two commands: documents, OCR text (no re-OCR), tags, correspondents, types, notes, custom fields. Storage paths become folders. Every file is checked against its SHA-256.
# 1. In Paperless (read-only export) docker compose exec webserver document_exporter ../export # 2. In Papiro docker compose run --rm -v /path/to/export:/import:ro \ papiro import /import
If you need those, Paperless is great. If you want something light, fast and easy to trust, try Papiro.
Yes. MIT-licensed, no paid tier, no telemetry, no account. You run it on your own machine.
On your server: one SQLite database and your original files in a folder. papiro export writes them out with readable names, so you never depend on Papiro to read your archive.
Images are published for amd64 and arm64. It idles at a few MB of RAM, so small boards and cheap VPSs are fine; OCR speed depends on the CPU.
No. Suggestions come from rules and from similar documents you already filed, computed on your server. An optional LLM integration exists, but it's off unless you add an API key.
Anything Tesseract supports. The image ships with English and Spanish; add more with one build argument.