This site gathers more than fifty open datasets about Sheffield, from Environment Agency rain gauges to Historic England listings, and stores a snapshot of each on a schedule. Where the licence allows it, that data goes straight back out again, free and without a key.
The reason is simple. Public information about a city should not end up locked inside one person’s website, and this is one person’s website. If it disappears, the useful part should not go with it.
Where to start
The catalogue lives at /wp-json/sheff/v1/data. It lists every provider with its licence, its attribution, its source, when it last refreshed, how many records it holds and whether it is redistributed. The sources page is the same information, readable.
Licences, and what is not served
Sources under the Open Government Licence, Creative Commons, the Open Database Licence or the Open Parliament Licence are passed back out. Sources with non-commercial or otherwise restricted terms are listed, so you know they exist, but not served, and the catalogue tells you why. Anything worked out here rather than fetched, such as the neighbourhood scores, is CC BY 4.0.
Every response carries its licence and a citation. CSV files carry them as comment lines at the top. If you use any of it, credit the original source rather than this site.
Getting the data
- One provider, as JSON, CSV, NDJSON or GeoJSON:
/wp-json/sheff/v1/data/deprivation?format=csv - A chart series:
/wp-json/sheff/v1/data/rainfall/chart/rain_daily - Everything about one neighbourhood in a single call, workings included:
/wp-json/sheff/v1/areas/crookes/data - All 147 neighbourhood scorecards at once:
/wp-json/sheff/v1/areas/scores?workings=1
Responses cache for an hour and are rate limited per address, which is plenty for research and for a dashboard that refreshes now and then. There is no key and no sign-up.
Widgets, and AI tools
The developers page has a widget builder: an area scorecard, a venue card, the fuel table, the storm overflow map or an events list, dropped into your own site by iframe or one script tag. The same data is exposed to AI assistants through a public MCP server with read-only tools, so a model can answer a question about Sheffield and say where the answer came from.
How it works underneath
Each source has a small connector that fetches on a schedule with a descriptive user agent, keeps the last good copy if a fetch fails, and never fetches while somebody is loading a page. Neighbourhood membership is point in polygon against the Data for Action boundaries. Anything derived is computed from the stored snapshots, and the inputs are always shown, because a number without its workings is just a rumour.
If you build something with this, I would genuinely like to see it. If you find an error, the sources page says where each dataset came from so it can be fixed at the root.