Curated, living catalogs of open government datasets across climate & energy and economy & small business, harvested from data.gov (US), data.gov.uk, Socrata portals (US & EU), Zenodo, Hugging Face, the U.S. Energy Information Administration (EIA), Federal Reserve Economic Data (FRED), and national open-data portals in Canada, Australia, Argentina, and Mexico.
Every entry is license-verified before publication — no fragmented portals, no license guesswork.
New datasets land regularly. Get the notable ones in your inbox instead of checking back here:
- Free weekly digest — the week's most notable new datasets, every Monday
3,927 license-cleared datasets, one row per dataset, in both CSV and JSON.
| File | Datasets | Covers |
|---|---|---|
climate-energy.csv / .json |
1,260 | Climate, energy, emissions, weather |
economy-small-business.csv / .json |
1,143 | Economy, business, employment, trade |
general.csv / .json |
1,524 | Harvested datasets outside the two verticals |
LICENSES.md |
— | License & attribution register — read before republishing |
export-manifest.json |
— | Generation metadata and counts |
The catalogs are also mirrored as Hugging Face datasets, with tabular preview and Hub search:
- GovDataHub/climate-energy-open-datasets — 1,260 datasets
- GovDataHub/economy-small-business-open-datasets — 1,143 datasets
- GovDataHub/general-open-datasets — 1,524 datasets
import pandas as pd
climate = pd.read_csv("climate-energy.csv")
climate[["title", "publisher", "license", "landing_url"]].head()# All CC-BY datasets in the economy vertical
csvkit climate-energy.csv 2>/dev/null || \
python3 -c "
import csv
for r in csv.DictReader(open('economy-small-business.csv')):
if r['license'] == 'cc-by':
print(r['title'], '|', r['landing_url'])
"Only datasets with allowlisted licenses are published: US public domain, public domain, CC0, CC-BY, Open Government Licence v3.0, and the Open Government Licence – Canada family (federal and provincial).
Datasets whose license could not be verified are excluded, never guessed — 2,239 harvested datasets are currently quarantined until their licensing checks out.
Some US rows carry license_source: inferred-federal: data.gov's API omits license fields, so these were classified by unambiguous federal authorship under 17 U.S.C. 105 (works of US federal employees). Contractor-produced data is excluded from this inference by design. See LICENSES.md for the full method and attribution requirements.
Attribution — OGL v3.0 datasets (most UK rows):
Contains public sector information licensed under the Open Government Licence v3.0.
Include the statement above (or a link to https://www.nationalarchives.gov.uk/doc/open-government-licence/version/3/) wherever you republish these catalogs.
Attribution — OGL – Canada datasets:
Contains information licensed under the Open Government Licence – Canada.
For federal Canadian datasets include the statement above (or a link to https://open.canada.ca/en/open-government-licence-canada) wherever you republish these catalogs; provincial datasets carry the equivalent provincial licence — attribute the province named in each row's publisher column.
The catalog above is free and stays free. Team Watch is the monitoring service built on the same pipeline, for teams whose products depend on specific datasets:
- Daily monitoring of your watchlist — new, changed, removed, or unreachable datasets flagged within 24 hours of detection, with provenance and timestamps.
- License-change alerts — a distinct alert class, because the license you cleared at integration time is not guaranteed to be the license in force today.
- Monthly license-clearance report — every dataset on your watchlist with its license, license basis, source record, and last check date.
- CSV/JSON change feed drawn from a timestamped change archive running since October 2026.
Pilot pricing: $99/month (launch $199/month; annual $1,990). Alert precision is measured and reported to pilot customers monthly; our operating standard is ≥90% of alerts rated relevant. Quality rules — allow-listed licenses only, zero license false positives, every claim traceable to a source record — are published in full at govdatahub.com, along with a public corrections log that includes our own errors.
Send up to 20 datasets your product depends on and receive a written report within 5 days: license status and basis, freshness, and link health for each. Free, no obligation.
Request a Dataset Risk Check — support@govdatahub.com
A fixed-scope audit of the dataset set you intend to ship commercially: provenance, license basis, and risk flags per dataset, in a written report. Quoted before work starts. A data-provenance audit, not legal advice.
- Custom dataset sourcing — $500–$2,500 per project. Tell us the dataset you need; we find it, verify the license, and deliver it normalized, with the license basis traced to the source record. Fixed quote before work starts.
- Point-in-time archive access — quoted per request. Dataset versions as published in prior periods, drawn from our timestamped change archive (running since October 2026).
- Due-diligence audits — $2,000–$5,000 fixed scope. For investors and acquirers: an audit of a target's open-data dependencies — provenance, license basis, and risk flags per dataset, in a written report. A data-provenance audit, not legal advice.
Details and pilot terms: govdatahub.com
This export was generated 2026-10-12. The pipeline re-polls its sources daily and audits resource links on a rotating sample; the catalogs on GitHub update with each published release (every release is human-reviewed before going out). Each row carries last_seen_at so you can check how fresh any entry is.
title, description, landing_url, portal_id, license, license_source, publisher, tags, vertical_suggested, last_seen_at
license_source—explicit(the portal stated the license) orinferred-federal(classified by federal authorship under 17 U.S.C. 105 — see License policy).vertical_suggested— a heuristic classifier label, not a human curation decision. Treat it as a starting point, not gospel.
GovDataHub (2026). GovDataHub — Open Government Dataset Catalogs.
Harvested from data.gov, data.gov.uk, Socrata, Zenodo, Hugging Face, EIA, FRED, and national CKAN portals.
https://github.com/GovDataHub/open-data-index
Machine-readable citation metadata lives in CITATION.cff. DOI (all versions): https://doi.org/10.5281/zenodo.23308617
Is this affiliated with data.gov or data.gov.uk?
No. GovDataHub is an independent curation project. Always verify current terms and availability at each dataset's landing_url.
Can I use these catalogs commercially?
The catalogs are a metadata index. Underlying datasets carry their own licenses (see LICENSES.md) — US public domain rows have no restriction; OGL and CC-BY rows require attribution as described above.
How is licensing verified? Explicit license fields from the portals where present; for US federal records where the API omits the license, classification by unambiguous federal authorship under 17 U.S.C. 105, with contractor-produced data excluded. Anything unverifiable is quarantined, never published.
How do I get notified of new datasets? Subscribe to the free weekly digest — the most notable additions, every Monday.
Suggestions, bug reports, and dataset requests are welcome.
- To report broken URLs or incorrect classification, open an issue.
- To suggest new portals or tags, submit a pull request.
- These catalogs are a metadata index, not a data mirror.
- Data-quality correction (2026-10-12): 205 datasets previously catalogued under the UK Open Government Licence v3.0 were found, on re-verification against the source APIs, to carry Canadian provincial open-government licences (Alberta, British Columbia, Nova Scotia, and others) or the federal Open Government Licence – Canada. They were removed from these catalogs pending licence allow-list review. UK OGL 3.0 entries are now sourced only from UK portals. Resolved later on 2026-10-12: the Open Government Licence – Canada family (federal and provincial) was allow-list approved, and those datasets are published again under their correct Canadian licences — 529 datasets in this release.
- Vertical labels are heuristic; license labels are the one field we never approximate without provenance.