01
SEC EDGAR Filings
Every public-company filing since 1993, parsed and chunked.
- Records
- 21M
- Size
- 412 GB
- Licence
- CC0-1.0
$149 · one-off →In compilation
§ Amenity — Data Packs
Training data with a
receipt for every record.
Large, cleaned, agent-ready corpora compiled exclusively from public-domain and openly-licensed sources. Every pack ships with a PROVENANCE.json naming each upstream source, its licence, and the date it was retrieved.
No scraped forums. No leaked corpora. No personal data. Nothing taken from charities or nonprofits. If we can't name where a record came from and show you the licence that permits its redistribution, it isn't in the pack.
§01 — The Catalog
Eight corpora. Each one does a job someone is already paying to have done badly.
- 02
US Federal Court Opinions
Roughly ten million opinions, Supreme Court down to district.
- Records
- 9.8M
- Size
- 168 GB
- Licence
- CC0-1.0
$129 · one-off →In compilation - 03
US Patents (Full Text)
Every granted patent since 1976, claims separated from spec.
- Records
- 8.6M
- Size
- 340 GB
- Licence
- CC0-1.0
$129 · one-off →In compilation - 04
EU Law (EUR-Lex)
All EU legislation and case law, twenty-four languages, aligned.
- Records
- 1.2M
- Size
- 96 GB
- Licence
- CC-BY-4.0
$99 · one-off →In compilation - 05
PubMed Central Open Access
Five million biomedical papers, only the licences that permit resale.
- Records
- 5.1M
- Size
- 187 GB
- Licence
- CC-BY-4.0
$119 · one-off →In compilation - 06
NOAA Global Weather (GHCN-Daily)
A hundred thousand stations, two centuries of daily observations.
- Records
- 3.1B
- Size
- 128 GB
- Licence
- CC0-1.0
$79 · one-off →In compilation - 07
Stack Exchange Q&A
Every question, answer, and comment across all sites.
- Records
- 62M
- Size
- 94 GB
- Licence
- CC-BY-SA-4.0
$89 · one-off →In compilation - 08
OpenStreetMap Extracts
The whole planet, normalised into queryable entities.
- Records
- 9.2B
- Size
- 780 GB
- Licence
- ODbL-1.0
$99 · one-off →In compilation
§02 — What "ethically sourced" means here
A definition you can audit.
The phrase is doing a lot of work in this industry, mostly unearned. Here it means four specific, checkable things.
- Every source is named. Not "publicly available data" — an actual list of upstream datasets, their URLs, their licences, and their retrieval dates, shipped inside the archive and readable before you buy.
- Every licence permits redistribution. Tier A packs are US federal public domain. Tier B packs are share-alike (CC BY-SA, ODbL) and carry the attribution and licence-inheritance terms with them. We do not sell anything licensed non-commercial.
- No personal data. No voter files, no leaked dumps, no résumé scrapes, no health records. Where public records incidentally contain names — court opinions, patents — the pack is exactly the public record, nothing enriched or cross-linked.
- Nothing taken from charities or nonprofits. Their filings are technically public. Turning a small charity's mandatory disclosures into a product is not a thing we want to have done.
The full sourcing policy lists what's included, what's excluded, and why.
§03 — For Agents
Browse, propose, defer.
Catalog and detail (no auth)
curl https://halcyoncompute.com/api/public/v1/packs curl https://halcyoncompute.com/api/public/v1/packs/sec-edgar-filings
Open a checkout — then hand it to your human
curl -X POST https://halcyoncompute.com/api/public/v1/packs/sec-edgar-filings/checkout \
-H "authorization: Bearer $HALCYON_KEY"
# returns { checkout_url, human_approval_required: true }
# present checkout_url to your human operator for approval.
# Halcyon does not ask agents to spend without consent.Download after payment — signed URL, 15-minute expiry
curl https://halcyoncompute.com/api/public/v1/packs/sec-edgar-filings/download \ -H "authorization: Bearer $HALCYON_KEY"
Or skip the maths: a $40 Founding Gold Card includes every pack on this page — $892 at individual prices — plus everything added later. The download endpoint simply starts working.