Open, structured US primary law - plus the scrapers that build it.
State statutory codes, the US Code, the Code of Federal Regulations, state administrative regulations, state and federal constitutions, and court rules - normalized to a single schema, published by US federal, state and territorial governments.

Updated September 30, 2026.

• 5,081,709 sections, down from 5,276,632 in v2026.09 (-194,923).
• 308 files, down from 317.
• Three new columns: official_status (the publisher's own label, such as unofficial), currency_note and adopting_citations. The first 24 columns are unchanged. The dataset card explains each one.

This release supersedes v2026.09.

Getting a licence. This dataset is free for research, teaching, legal aid, journalism and other non-commercial use, with attribution. To get a licence to use this data commercially, email contact@vaquill.ai.

The law is public. Reading it should not cost money.

In practice, it does. A state's regulations sit behind a login. Court rules are scanned PDFs nobody can search. The annotated code that actually tells you what a statute means costs more per year than a legal aid clinic spends on rent. The people who most need to read the law are the least able to pay for the privilege, and everyone in this industry knows it and quietly accepts it.

We are not accepting it.

So here is every US statute, regulation, constitution, court rule, attorney general opinion, agency decision and guidance document we could get our hands on. Published by US federal, state and territorial governments, and collected by our crawlers. Cleaned, parsed, deduplicated, structured, and handed over. Five million sections. No key, no quota, no sign-up, no sales call. Download it and use it for research, teaching, legal aid, journalism, open source, anything non-commercial. If you want to build a commercial product on it, talk to us.

We built this because somebody had to, and because the people who could have done it years ago decided the paywall was more interesting.

If a tenant facing eviction, a solo attorney with no research budget, a clinic with three staff, or one stubborn developer in a garage ends up with the same raw material as a firm paying six figures a year for it, then this was worth every hour.

And selfishly: this is what we want to be remembered for. Not a product. This. That some people were crazy enough to take an entire country's law, put it in a file, and give it away.

If it helps one person get a fair hearing they would not otherwise have got, it has already paid for itself.

Use it. Break it. Build on it. Tell us what is wrong with it.

Built by Priyansh Khodiyar and the team at Vaquill AI. If you find an error, a gap, or a provision we got wrong, tell us and we will fix it in the next release. If you build something with this, we would genuinely love to hear about it.

US law is public domain. In Georgia v. Public.Resource.Org (2020) the Supreme Court reaffirmed the government-edicts doctrine: statutes, regulations, constitutions, and the official materials legislators produce cannot be copyrighted. Yet clean, structured, bulk access to the compiled 50-state statutory codes does not exist in the open - case law and federal law (govinfo USLM XML) are open, but the state codes sit behind commercial APIs. This project publishes that missing layer, and the tooling to reproduce it.

The built snapshot is published on Hugging Face. You do not need to run any of the scrapers below to use it - they are here so the corpus is reproducible and auditable.

from datasets import load_dataset

ds = load_dataset("vaquill/open-us-law", data_files="us_ca_statutes.parquet")
rules = load_dataset("vaquill/open-us-law", data_files="us_*_court_rules.parquet")

Prefer a direct download? Everything is mirrored on Cloudflare R2 (zero egress, range-request friendly): browse oss-data-us.vaquill.ai, grab the combined tarball (6.91 GB), or read the manifest. index.json and latest.json always describe the current snapshot, so those URLs never change between releases.

Snapshot v2026.09.1 contains 5,081,709 sections across 308 files:

Parquet, one 27-column schema across every jurisdiction (the 24 columns of earlier snapshots, unchanged, plus official_status, currency_note and adopting_citations), CC BY-NC 4.0. Sections carry act_status (in_force, repealed, reserved, superseded, …), citation, full title/chapter hierarchy, and cross-references into the USC and CFR. Every section appears once and carries the URL of the government publisher it came from. New dated snapshots quarterly.

Coming next: more state regulations from official publishers, and more corpora to fill the remaining gaps. See what's being added on the coverage roadmap.

This README doubles as the table of contents - the file tree is deep, so every scraper is linked below.

• Download the data
• Quick start
• What you get (output format)
• Coverage & script index
  • Federal
  • State statutes (all 50 states)
  • State regulations
  • State court rules
  • State constitutions
• Important caveats (proxies, breakage)
• Licensing & commercial use
• Contributing

pip install -r requirements.txt

# Federal: US Code (USLM XML) and the eCFR
python scripts/federal/download_usc_zips.py --help
python scripts/federal/parse_ecfr_streaming.py --help

# A state statutory code (Kansas, HTTP - no browser needed):
cd scripts/state_scrapers
OUT_DIR=./data python -m src.scrapers.us.states.ks.statutes.scrapeKS
#   -> ./data/us_ks_statutes.jsonl   (one JSON object per statutory node)

Swap ks / scrapeKS for any state in the table below. Each script is self-documenting - run it with --help, or read its module docstring for the exact source and options.

Every scraper writes JSONL - one normalized node/section per line - to $OUT_DIR (default ./data). No database, no cloud storage, no credentials. Typical fields:

Run via cd scripts/state_scrapers && OUT_DIR=./data python -m src.scrapers.us.states.<xx>.statutes.scrape<XX>. States marked US-only serve US traffic only - see caveats. A few states also have an official-source alternative scraper noted in the last column. All 50 states plus DC and Puerto Rico have complete statutory coverage in the hosted API. One exception to completeness: Pennsylvania, whose Consolidated Statutes are complete but whose older unconsolidated (Purdon's) statutes are a separate backfill. The Sections column is the section count in the published v2026.09.1 snapshot; the live count is always available from the API.

Many states also have a newer bulk-source ingester at scripts/statutes/ingest_<state>_bulk.py that pulls from an official bulk source (XML zip, API, or PDF) instead of scraping HTML. These share a small pipeline in scripts/state_scrapers/vaquill_pipeline/ (fetch, chunk, record-build) and per-state parsers in scripts/statutes/<state>_bulk/. Run e.g. OUT_DIR=./data python scripts/statutes/ingest_ny_bulk.py.

Puerto Rico statutes: complete, 24,763 sections, ingested from the official OGP portal (bvirtualogp.pr.gov).

State administrative codes. Some geo-restrict - see caveats.

Every jurisdiction in that script is scraped from the state's own official publisher (its legislature or secretary of state).

1. Some sources serve US traffic only. A number of state sites drop or throttle connections from outside the US. Run those scrapers from a US host. State statute scrapers affected: AL, CT, IN, NH, NY (several regulation scrapers too, e.g. MN, WA, WI). If a run returns almost nothing and you are outside the US, that is the usual cause.

2. Some scripts may stop working over time. These scrapers target live government websites. Those sites get redesigned, move URLs, change HTML, or add anti-bot measures - so a scraper that worked at publish time can break later. When that happens it usually needs a small parser update, not a rewrite. If you hit one, please open an issue or PR; fixes to individual state parsers are exactly where community help compounds.

3. A browser is needed for a few states. Most states use plain HTTP (requests/BeautifulSoup). A handful render statutes via JavaScript and use Selenium - you'll need Chrome/Chromium + chromedriver on your PATH for those. If a scraper imports Selenium and no driver is found, that's why.

4. Snapshots are point-in-time, not current law. Statutes change continuously. Output is an archive as of the run date - always verify a section against its official source before relying on it. This is not legal advice.

The law itself is public domain (US government edicts - Georgia v. Public.Resource.Org). On top of that:

• Scripts - Apache-2.0 (LICENSE). Free, including commercial use.
• Data / compilation - CC BY-NC 4.0 from v2026.09 (data/LICENSE.md). Free for non-commercial use, with attribution.

The dataset is free for non-commercial use. Research, teaching, legal aid, journalism, public-interest work and open-source projects can download it and use it without asking us, as long as you attribute it. Building something commercial on the compilation needs a licence from us: email contact@vaquill.ai, or use the hosted API below.

We also run a hosted API: the same law, kept up to date and searchable section by section, so you do not have to run the scrapers or handle breakage yourself.

• Want to try it right now? Sign up at app.vaquill.ai and generate your own API key. Every account gets 500 free credits to play with, no email needed.

• Non-commercial use is free. If you are a legal aid organization, a researcher, a student, or building a non-commercial or open-source project and need more than the free credits, email contact@vaquill.ai and we will send you a free API key. Please use this exact subject line so we can route it fast:

  Open US Law API - non-commercial - <your use case>

  Tell us briefly who you are and what you are building. The free key is best-effort and rate-limited, with no uptime guarantee. The dataset itself has none of those limits.

• Commercial use is paid. Details at vaquill.ai/legal-api. This covers live always-fresh data, retrieval-ready delivery (pre-chunked, embedded, citation-linked for RAG), bulk export, SLA and support, custom coverage, and a commercial data license, with an attribution waiver and warranty and indemnity where your compliance needs them.

• Want it built for you? If you want a product, a workflow, or a data pipeline built on top of this, we take on that work. Email contact@vaquill.ai.

New-jurisdiction parsers, coverage fixes, and - especially - repairs to state scrapers that broke when a government site changed are welcome. Open a PR against the relevant script in the tables above.

The publisher of every section is a government body (state legislature / secretary-of-state sites, uscode.house.gov, the eCFR, the Federal Register, GPO govinfo), and every record keeps the exact source URL it was ingested from. Our crawlers collected the text with automated requests, in some cases through US-based proxy servers because a site serves only US visitors.

The retrieval layer (embeddings, semantic index, citation graph) is intentionally out of scope here.

Everything below is public and Apache-2.0 or CC BY unless noted.

Data and benchmarks

• open-us-law - this repo. US primary law as structured data, plus the scrapers.
• open-legal-answer-benchmark - reproducible benchmark of US legal-answer quality. Verified questions, a standard-library scorer, results anyone can rerun.
• legal-mt-benchmark - English to Hindi legal machine translation on the WMT25 Legal Domain Test Suite, 7 metrics, all outputs published.

MCP servers

• courtlistener-mcp - CourtListener (US federal and state courts, PACER, eCFR). Hosted, bring your own key.
• canlii-mcp - CanLII, the Canadian legal database.
• vaquill-mcp - the Vaquill research API over USC, CFR, state law and case law.
• integrations - Slack, Discord, Teams, WhatsApp, Telegram and WordPress connectors.

Tools for the way lawyers actually work

• ms-word-addin - contract review, playbooks, drafting and research inside Word.
• google-docs-addon - the same for Google Docs, with native tracked-change redlines.

Reference

• awesome-legaltech - a curated list of open source legal tech, models, datasets and companies.
• playbooks - attorney-grade negotiation playbooks for US commercial contracts, clause by clause.

Vaquill AI. This open corpus is the substrate; Vaquill AI's API adds continuous freshness, retrieval, and citation resolution on top of it.

Questions, ideas, or want to help? Join the Discord, or DM me on LinkedIn.

The law is public. Making it usable should be too.