Barcelona

No ads · No trackers

A software company in Barcelona, publishing its own accounts

A3The Morgue

Web archiving, run for people who cannot run it

Archivist

Preservation-grade web archiving with an editorial workflow, bilingual search, and library-standard interoperability — hosted, so a three-person organisation does not have to become an infrastructure team.

Newspapers called it the morgue: the room where every clipping was filed, indexed, and kept, so that a claim made today could be checked against what was actually printed a decade ago. It was unglamorous, expensive, and the single most valuable thing in the building — because an institution that cannot check its own past has to take somebody's word for it.

Most organisations with a record worth keeping now have nothing of the sort. The record lives on the web, which forgets: links rot, platforms close accounts, publications quietly revise. The tools that would fix this exist, but they are national-library infrastructure — crawlers, WARC, fixity auditing, persistent identifiers, harvesting protocols — and no three-person memory project is going to stand that up between grant applications.

Where it came from

Every part of this was built and proven for July Archive, which preserves the primary record of Bangladesh's July 2024 revolution. That archive is free, non-profit, and stays that way. This is the same machinery offered to organisations that need it and would otherwise go without.

The reference deployment

July Archive

Preserving the primary record of Bangladesh's July 2024 revolution

A bilingual (English/Bengali) public archive of the 36 days that toppled a regime — documents, photographs, testimony, and broadcasts, preserved as primary sources with citations. Open API access for researchers, and answers drawn only from the published record.

julyarchive.org

What it does

01

Capture, not screenshots

Pages are recorded as WARC — the format national libraries use — so what is preserved is the page as it was served, replayable years later, not a picture of it. Scoped, polite crawls that honour robots by default, and a browser-based capture path for the pages that only exist once JavaScript has run.

02

An editorial gate

Nothing is published because a crawler found it. Captures land in a review queue where a person checks provenance, rights, and sensitivity before anything becomes public — with an append-only record of who decided what, and when.

03

Findable in every language you hold

Search and question-answering that work across languages, so a question asked in one surfaces sources written in another — with answers grounded only in the collection and cited back to the record, never invented.

04

Built to be cited

Persistent identifiers that keep resolving, checksums audited on a schedule, and OAI-PMH so libraries and aggregators can harvest you into their catalogues. An archive nobody can cite or find is a hard drive.

05

Withdrawal that actually withdraws

A takedown removes the record, its captures, and its search entries, and propagates the deletion to anyone harvesting you — because the right to be forgotten is worthless if it only hides a row.

06

Not a hostage

Export at any time in open formats. If you outgrow us, or stop trusting us, you leave with your collection intact. Preservation that depends on one vendor's survival is not preservation.

Terms, plainly

We do not take ownership of your record

Hosting a collection gives us no rights over it. Rights in the material stay where they were — with the people and publications that made it — and nothing in it is used to train models, ours or anyone's.

Leaving does not cost you the archive

Captures, metadata, and identifiers export in open formats whenever you ask. Preservation that only survives while one supplier does is not preservation, and we would rather say so than discover it later together.

These sit alongside the company's other dated promises in Standards & Practices, page B2.