Docspell

Document organizer with OCR, NLP metadata suggestions, HOCON config, a full REST API, and the dsc CLI.

AdaptableOperable Business appWeb app Deterministic ScalaElm API License: AGPL-3.0

Docspell ingests scans, email attachments, and uploads, runs OCR, suggests correspondents, tags, and dates with Stanford CoreNLP, and stores everything with custom fields and full-text search. It targets households and smaller companies. Two Scala services (restserver and joex) are configured with HOCON files or env vars; every feature is reachable over the REST API, which the dsc CLI and Android app use.

Use it when

we scan supplier invoices and receipts and want them OCR'd, dated, and matched to the right vendor automatically before the bookkeeper sees them
we want our agent to push documents in through the integration endpoint and search them by vendor and date range

Gotchas

AGPL-3.0 or later
JVM plus tesseract, ocrmypdf, unoconv; the joex worker is memory-hungry with NLP enabled
Session tokens default to 5 minutes; long agent sessions must renew through /sec/auth/session
Positioned for home and small groups; multi-tenant is by collective, not by organization hierarchy

Last activity Sep 2, 2026. Fit: adaptable 2, operable 3.

Editions

Free (self-hosted) · Free open source

Clonegit clone https://github.com/eikek/docspell
ImageContainer
DownloadDownload

For agents: JSON · Markdown

Adaptable: what is yours to change

Make it yours

Config only9 things to edit

Two HOCON files (docspell.server.* and docspell.joex.*) or env vars in the DOCSPELL_SERVER_APP__NAME pattern. An SMB sets docspell.server.base-url, backend.signup.mode (closed after the first collective), backend.jdbc, full-text-search backend, joex OCR languages and ocrmypdf options, the integration-endpoint secret for scanner uploads, and SMTP for outgoing mail. Correspondents, tags, custom fields, and scan-mailbox jobs are created in the app.

Env vars: DOCSPELL_SERVER_APP__NAME DOCSPELL_SERVER_BASE__URL DOCSPELL_SERVER_BACKEND_SIGNUP_MODE DOCSPELL_SERVER_BACKEND_JDBC_URL DOCSPELL_JOEX_BASE__URL
Config files: docspell-server.conf docspell-joex.conf modules/restserver/src/main/resources/reference.conf modules/joex/src/main/resources/reference.conf
Prompt files: none
Accounts you need: none; optional SMTP and IMAP mailboxes

Source

github.com/eikek/docspell
AGPL-3.0 license
Last commit Sep 2, 2026
2,317 stars

Operable: what an agent can drive

Interface

OfficialAPIpublished by the builderlocal

How an agent can drive an Operable Listing, in five levels: native (built for the agent as the primary interface), official (the Builder ships an agent interface), wrapped (a third party wraps the software's API), config (an agent can only drive it by editing files), gui (screen control only). An interface is official only when it is published by the same organization that ships the software. GUI-level Listings are allowed and sorted last.

Docs: docspell.org/docs/api/

Reach

Partial coverage. 9 of 10 jobs reachable through the stated interface.

Job	Through the interface	Evidence
Upload documents	Reachable	docs/api: upload request; integration endpoint and source URLs for unauthenticated uploads; dsc CLI watch and upload
Search with the query language	Reachable	docs/api/intro: protected route /sec/item/search
Edit tags, correspondents, custom fields, dates	Reachable	README: everything is accessible through a REST/HTTP api; web app is a client of it
OCR and text extraction	Reachable	joex processes every upload; no action needed beyond config
Metadata suggestions from NLP and learned classifiers	Reachable	README: suggests correspondents, guesses tags, finds dates using machine learning; runs automatically
Import from mailboxes	Reachable	README: great e-mail integration; scan-mailbox tasks configured through the API-backed settings
Full-text search across content	Reachable	README: fulltext search; admin route to rebuild the index
Collective insights and statistics	Reachable	docs/api/intro: /sec/collective/insights
Send items by mail or share them	Reachable	README screenshot mail-item; share feature documented on docspell.org
Book-keep amounts or reconcile	Not reachable	archive only; custom fields can hold amounts but there is no ledger

Code stack

Languages: ScalaElm
Frameworks: http4sdoobiefs2Stanford CoreNLPTailwind
Package manager: sbt
Database: PostgreSQLMariaDBH2
Hosting target: Docker Compose (docspell/docker)deb packagezipNix / NixOS moduleHelm chart
Built: Human-written (recorded, never ranked)

Evidence and confidence

Field	Where it was seen	Confidence
formFactor	README: SPA web app over a REST server; docker-compose quick start on port 7880	95%
executionModel	Stanford CoreNLP classical NLP, no LLM calls	90%
operability	README: everything is accessible through a REST/HTTP api; docs/api/intro documents auth, search, admin routes; dsc CLI and Android app are clients	90%
effort	docspell.org/docs/configure/main: HOCON config, env var mapping, docspell.server and docspell.joex sections; reference.conf in repo	85%
editions	README: free software under AGPLv3; no paid edition	95%
active	gh api: pushed 2026-09-02, 2317 stars	–
reach		85%

Open questions

Whether any MCP wrapper exists for Docspell (none surfaced in sourcing)

Sources

github.com/eikek/docspell sourcing:support-bookkeeping:curated
docspell.org/docs/api/ sourcing:support-bookkeeping:curated
docspell.org/docs/configure/main/ analysis