Docspell is a self-hosted document management system that helps households, families, and small groups or companies organise the piles of paper, scans, and e-mails that accumulate from scanners and other sources, using machine learning and OCR to cut down the manual effort of adding metadata.

What it is

Docspell is a personal document organiser, also described as a Document Management System (DMS). It is written primarily in Scala with an Elm front end, is licensed under AGPL-3.0, and lives in the open-source document management ecosystem, alongside the topics dms, edms, ocr, nlp, and personal-document-system. It takes files arriving from scanners, e-mails, and other sources and unifies them into one searchable store. Documents can carry tags, correspondents, dates, and other predefined or custom metadata, and everything is reachable through a REST/HTTP API.

The concrete problem it replaces is the routine of manually filing and labelling scans. Adding metadata by hand is tedious, so Docspell suggests correspondents, guesses tags, and finds dates using machine learning. It learns metadata from the documents already present and locates things using NLP, backed by the free GPL-licensed Stanford CoreNLP library. Where needed it runs OCR on documents, provides fulltext search, and offers e-mail integration — replacing the loose pile of scanner output and mailbox attachments that would otherwise have to be sorted by hand.

Key capabilities

• Automatic metadata assistance: suggests correspondents, guesses tags, and finds dates using machine learning trained on existing documents, with NLP via Stanford CoreNLP.
• OCR processing on documents when the uploaded file requires it, feeding fulltext search.
• Fulltext search over the organised store, once metadata and extracted text are associated.
• E-mail integration for pulling documents in from mailboxes.
• REST/HTTP API exposing the whole system, with a mobile-friendly SPA web application as the default user interface.
• Android app for uploading files from a phone or tablet, plus a command-line client, dsc.
• Metadata model covering tags, correspondents, and both predefined and custom fields.

Who uses it and how

• Home use: families and households scanning paper correspondence and filing it into one collective.
• Smaller groups and companies that need a shared DMS without a hosted service.
• Users who already own a scanner and want the resulting files organised with minimal manual tagging.
• People who receive documents by e-mail and want them captured automatically rather than saved by hand.
• Mobile-first users who photograph or share documents from a phone or tablet through the Android client, then curate them later in the web application.

Getting started

The quickest route is the Docker setup from the docspell/docker repository: clone it, enter docker-compose, run docker-compose up -d, then open http://localhost:7880, sign up, and log in using the same name for collective and user. Alternatives are documented on the get-started and quickstart pages: installing the provided deb file on a Debian-based system, downloading the provided zip and running the script in bin/, or using the nix package manager.

How it compares

No comparable tools are named in the supplied facts, and no list of paid products this project replaces is provided. On the evidence available, Docspell stands alone in this registry. Its AGPL-3.0 licence, its self-hosted multi-component architecture, and the absence of any hosted or paid tier mean the comparison has to be made against whatever else an operator is already running, not against named commercial alternatives.

When to use it — and when not to

A self-hoster must run and maintain the whole multi-component stack that the docker-compose setup brings up, keep up with updates, and configure the e-mail integration against real mailboxes, since mail handling is a core rather than optional feature. It is a poor fit for anyone who wants a managed service, for organisations that cannot accept AGPL-3.0 copyleft obligations, or for teams unwilling to operate a database-backed server alongside a browser client. The repository itself carries 235 open issues, which is worth weighing as a signal of maintenance load, and the machine learning component depends on an external GPL library, so the dependency chain is worth reviewing before standardising on it.

git clone https://github.com/docspell/docker docspell-docker
cd docspell-docker/docker-compose
docker-compose up -d