About the Toolbox

A card catalogue of what historians actually used.

The Digital Historian’s Toolbox collects the software and public datasets historians work with, and links each one to the projects that put it to use. It is community-submitted and editor-reviewed.

Why it exists

General tool directories already exist, and they’re useful. But they cover every humanities discipline, and they mostly tell you what a tool is, not what happened when a historian used it. Meanwhile, new AI tools arrive faster than anyone can test them on real archives.

The Toolbox is narrower on purpose. It is built for historians, it records real uses with links you can check, and it is frank about how tools fail on historical sources.

Research is not the same as a benchmark

A benchmark tells you how well a tool reads. It doesn’t show that a historian learned anything with it. So every card sorts its projects into three kinds:

Research: The tool or dataset helped answer a historical question: an article, a book, a project that makes an argument about the past.
Corpus building: It produced something other historians can use: transcriptions, captions, a gazetteer, an enriched or re-OCRed dataset.
Evaluation: The tool itself was tested: benchmarks, accuracy studies, feasibility case studies. Valuable, but it isn’t use.

Historian’s notes

Every tool card answers three questions: where do my sources go when I use it, how has it been tested, and how does it fail on historical material. Dataset cards instead say how the text was made, what gaps and biases it carries, and what you’re allowed to do with it.

These notes matter most for AI tools, where hallucinated text, silent spelling “corrections” and unclear data terms are common and easy to miss.

How sorting works

Trending counts research and corpus-building uses within a time window you choose, plus “I’m using this now” check-ins from historians who are signed in. Evaluations never count, so a burst of benchmarks can’t push a tool up. Most used counts all time, which favours older tools; Newest and A–Z do what they say.

Editorial policy

Every submission is reviewed by an editor before it appears.
Every project link is opened and checked to confirm it actually mentions the tool or dataset.
Anything unverified or contradictory is flagged on the card, never quietly filled in.
Conflicts of interest are disclosed on the card, including the editors’ own publications.
Every card shows when it was last updated. Cards not checked in 12 months are marked “Due for review”, because prices, licences and data terms change.
Anyone can suggest a correction from any card, and each card keeps a version history.

Built to outlast its editors

Tool directories have a habit of disappearing when their volunteers run out of time; DiRT, the Digital Research Tools directory, survived only by being folded into TAPoR. The Toolbox is built so its contents outlive both its editors and its website:

Open data on GitHub. Every card, project link and historian’s note is published as plain JSON and CSV files that anyone can download, reuse or fork, under a CC BY 4.0 licence. See the data.
Archived on Zenodo. Each release of the data is deposited in Zenodo, CERN’s research repository, with its own DOI, so a citable copy remains even if this site goes offline.
A small team of co-editors, so the catalogue never depends on one person’s time.

Your privacy

Submissions: your name is credited on the card if you give one; your email is used only by the editors to ask about your submission and is never shown.
“I’m using this now”: you sign in on orcid.org, so the Toolbox never sees your password. It stores your ORCID iD and name and which cards you checked in to, once per year. Only totals are ever shown, never who.