Skip to main content
5M+ distinct sections, indexed as 19M+ retrieval passages, across federal law and all 52 states and territories. The corpus grows continuously, so this page gives magnitudes and the table below gives the measured split. For a number you can put in a contract, call the endpoint: it is free, it is self-describing, and it is authoritative over anything written here by hand.

Federal

1.5M+ sections, 14 corpora. US Code, CFR, Statutes at Large, Federal Register, adjudications, and per-corpus scope limits.

States and territories

3.5M+ sections across 52 jurisdictions. Statutes, regulations, constitutions, court rules, state agency guidance and AG opinions.

Agency guidance

50K+ documents from 53 named federal sources, each filterable with source.

​
What’s in the corpus

The only place on this page that states exact counts, so there is one number per fact rather than the same fact restated in three places. Measured 2026-09-11; the endpoint above is live. Every section is indexed for hybrid semantic and keyword search, and mirrored in multiple formats (HTML, PDF, XML, plain text, DOCX) where the source provides them.
Counts here are distinct sections, one per provision or document, matching what GET /us/statutes/coverage returns. Longer documents split into several passages for retrieval, so the index holds more entries than sections, over half of them Federal Register rules at 13 to 15 passages each. That does not change what you get back: a citation lookup still resolves to one section with its full text.

​
Federal

14 corpus types. The US Code and the Statutes at Large, the CFR both as a live compilation and as 31 published annual editions, Federal Register rules and a curated slice of its notices, Executive Orders, agency guidance from 53 named sources, seven bodies of administrative adjudication, the Federal Rules, the Constitution, the Sentencing Guidelines, tax treaties and GovInfo statute compilations. Full federal table with per-corpus scope limits → Agency guidance is the one corpus with a registry behind it: 53 named sources across 26 agencies behind the single corpusType=AGENCY_GUIDANCE token, each narrowable with source. Source registry →

​
State and territory matrix

All 52 jurisdictions carry statutes and regulations. 51 carry constitutions and 50 carry court rules; the two gaps are source problems, not coverage decisions, and each is explained. Per-jurisdiction counts for all 52 →

​
State agency guidance

State-issued regulatory guidance that interprets state law, almost all of it insurance-regulator bulletins, across most jurisdictions. Queryable with corpusType=STATE_AGENCY_GUIDANCE paired with state. The count is in the table above. Unlike statutes and regulations this is not a comprehensive rollout: it lands state by state, and some states may never be built. Per-state sources and counts →

​
State Attorney General opinions

Formal written opinions of a state Attorney General, construing that state’s own statutes, constitution and regulations for a state officer, legislator or local government who asked. Queryable with corpusType=STATE_AG_OPINION paired with state.
These are persuasive authority, not binding law. Officials follow them in practice, and an AG opinion is very often the only interpretive authority that exists for a provision no court has reached. Every section in this corpus is served with goodLawStatus: "unknown" by construction: only three of the 52 jurisdictions publish any supersession signal at all, so we do not certify an opinion as current. See Status and currency.
Like state agency guidance, this is not a comprehensive rollout: it lands jurisdiction by jurisdiction, and it is early. The 52-jurisdiction survey behind this corpus measured roughly 184,600 individually addressable opinions, so what is live today is a low single-digit percentage of what exists. Use GET /us/statutes/coverage for the jurisdictions and counts that are actually live right now rather than relying on this page, which is written by hand and lags the rollout. Delaware is documented below because it is the reference build and because two things about its composition would otherwise surprise a buyer. It is not the only jurisdiction live. Delaware, 843 opinions, 1995-2026. Two things about its composition are worth stating plainly, because a buyer would otherwise be surprised by both:
  • Delaware runs one numbered series and does not separate its opinion types. Only about 17 of the numbered items are substantive Attorney General opinions; the rest are FOIA opinion letters interleaved in the same numbering. Both are formal opinions of the Attorney General under Delaware law, so both are here, but if you are looking for statutory interpretation you will mostly find open-records adjudication.
  • 173 of the 843 are served as the state’s letter recovered from a reporter print. Delaware publishes those posts in a commercial reporter’s rendering, wrapping the Attorney General’s own letter in that publisher’s citation and page-break marks. The opinion underneath is a government work, so we remove the wrapper and serve the letter; we do not reproduce the reporter’s citation or its pagination. 150 of the 173 exist in no other form in Delaware’s archive and span 1995-2011, so the alternative was to publish no Delaware opinion older than 2011.
  • 50 further posts carry no opinion number and are excluded: press releases, job postings and notices that share the archive but are not opinions.
Montana, 3,527 opinions, volumes 15-44 (roughly 1933-1992). Documented here because what is served is a deliberate subset and the reason is worth stating. The date range is DERIVED from the latest year appearing in each opinion’s text: Montana publishes no structured issue date for these, so they carry no year and are not reachable by a year filter.
  • Montana’s Judicial Branch publishes 8,516 opinion PDFs spanning 1899-2024. Volumes 1 to 14 are cited by PAGE, not by opinion number, and the publisher prints two opinions on one page, so there is no per-opinion identifier to key on. Those 3,969 are not served.
  • The per-opinion PDFs are page extracts, so an untrimmed document carries the opinion that follows it. Each body is cut to its own OPINION NO. header; the 1,010 where that header cannot be read, or where the opinion’s own continuation is not in its PDF, are withheld rather than served with a neighbour’s text under the wrong citation.
  • Opinions from volume 45 onward (1991 to date) are published by a different Montana office and are not included here.
Arkansas, 10,295 opinions, 1984-2026. The largest jurisdiction in this corpus. Two things to know:
  • 334 further opinions are listed by the Attorney General but not served by it. Arkansas’s own index carries a row for each, and the document behind it has never been published. Those are recorded as a coverage limit rather than counted, so the 10,295 is what we actually hold. They cluster in 1985-1998.
  • Six opinions run to 100,000 characters or more. Those are ballot-title certifications, where Arkansas requires the Attorney General to reproduce the entire proposed initiated act or constitutional amendment inside the opinion, and they are served whole.
Indiana, 9,759 opinions, 1906-2026. The deepest archive here, and the one with a limitation worth stating plainly:
  • Indiana publishes in two citation forms, and both are carried as the state prints them: numbered opinions (2026 Ind. Op. Att'y Gen. 02) from 1944, and page cites into the annual and biennial report volumes (1909-1910 Ind. Op. Att'y Gen. p. 284) before that.
  • 🔴 In the scanned eras, the state cut its bound volumes into per-opinion PDFs by page, so an opinion that runs past its page cut is severed in Indiana’s own file. We serve the full text of every document the state publishes, but for those opinions that text is the opening pages rather than the whole opinion. It affects roughly a quarter of the pre-1944 page cites and a third of the 1940s numbered opinions; the 1960s, 1970s and 2000 onward are unaffected. The continuation is usually present in the adjacent page cite, under its own citation.
  • Every opinion also carries the statutes it construes, the Attorney General who signed it, and Indiana’s own summary, all from the state’s published catalogue rather than mined from the text.
Idaho, 151 opinions, 1985-2026. Idaho publishes Guidelines, Certificates of Review, Annual Reports and Tables of Citations in the same feed as its opinions. Only the opinions are here; the other 602 documents are a different instrument and are excluded. Opinions from 1881-1984 exist only inside annual report volumes and are not yet extracted. Oklahoma, 68 opinions, 2023-2026. This is deliberately a 2023-onward corpus. Oklahoma’s older opinions are published on a network whose publisher does not permit automated access, so we do not hold them and do not intend to without permission. Florida, 20,923 opinions, 1913-2026, and the two halves are NOT the same product. Florida is the deepest AG archive we carry and its publisher serves it in two forms:
  • 1971-2026, 4,166 opinions, full text. The state publishes these as HTML with its own number, issue date and subject, and they are indexed and retrievable like any other opinion.
  • 1913-1991, 16,757 opinions, citation and metadata only, with NO opinion text. The state publishes this era only as scanned pages of the bound Biennial Report of the Attorney General, and those page images carry several opinions each: measured across the archive, 27% of them hold two or more opinions and only 55% print their own opinion number, so no per-opinion text can be extracted from them without splitting the volumes. Rather than serve one opinion’s number over another opinion’s words, these records carry the citation, the publisher’s own topic and a link to the publisher’s page. Every one of them is marked textAvailability: "metadata_only", so you can filter them out or in; the full-text records are marked "full".
Two further limits worth stating. 80 opinions are withheld: 44 because Florida gave two genuinely different opinions the same citation and we will not guess which is which, 32 whose printed citation we cannot parse, and 4 whose citation prints no date. And the state’s own archive lists no opinions at all for 2024, with four in 2023 and four in 2025; whether none were issued or none are listed is not something the archive answers. Kentucky, 692 opinions, 1992-2026. Every year from 1992 onward is represented. Kentucky’s earlier opinions were never published online, and the Attorney General’s office supplies them by phone or mail on request, so we do not hold them. Maryland, 601 opinions, 1993-2025. Two series, both carried as the state publishes them: 565 published opinions, cited by volume as 110 Op. Att'y Gen. 3, and 36 unpublished ones that the office numbers 00-123 and posts separately. Opinions from 1916 to 1992 exist only inside 77 bound-volume compendia, each a single file holding a whole volume rather than one opinion. Those volumes are not split into individual opinions and are not included. Puerto Rico, 81 opinions, 2003-2010, in Spanish. The text is Spanish as published, and language_code is es. The served range ends in 2010 because the opinions from that year onward are posted as page images with no text layer: 50 documents we can see are excluded until they are processed, rather than absent from the office’s own archive. Where a jurisdiction runs more than one series, the series field carries the publisher’s own label so you can filter them apart — Washington’s formal AGO against its AGLO letter opinions, Georgia’s Official against Unofficial.

​
Refresh cadence

How often each corpus is re-pulled from its official source. Counts grow between updates, so call GET /us/statutes/coverage for a live figure. For the age of any single source, call GET /boards. It is free and carries lastRetrievedAt, cadence and retrievalStatus per source. GET /us/statutes/coverage reports what each corpus is current through in the publisher’s own words, which is a different question and is not available for every state.

​
Which statute cadence applies where

Every 7 days, 25 jurisdictions. California, Colorado, the District of Columbia, Florida, Idaho, Illinois, Indiana, Iowa, Louisiana, Michigan, Mississippi, Missouri, Nevada, New Jersey, New Mexico, New York, North Carolina, Oregon, Pennsylvania, Puerto Rico, South Dakota, Utah, Virginia, Wisconsin and Wyoming publish their code in a form that can be pulled whole, an official bulk export, an API, or a complete per-section fetch. Every run re-pulls the entire code and reconciles it against what is stored rather than checking only what looks new. Every 91 days, 3 jurisdictions. Arkansas, Georgia and Tennessee are pulled the same whole-code way, on a quarterly slot rather than a weekly one, tracking legislatures that sit for part of the year. They report cadence: "quarterly" on GET /boards. Everywhere else, statutes are re-pulled monthly or carry text as of their ingest date. That is a fact you can read rather than assume: GET /boards returns lastRetrievedAt, cadence and retrievalStatus per source, and anything we have deliberately stopped re-pulling is declared paused in freshness with its reason. A cadence is a schedule, not a promise that the last run worked. Some boards are failing or have never completed a run at any given moment, which is why every statement above is worth reading off GET /boards rather than off this page. A failing board does not mean the corpus is gone: the sections stay queryable at the currentThrough the coverage response reports, they are simply not being re-pulled on schedule until it clears. Seven jurisdictions are covered but not watchable: Delaware, the District of Columbia, Hawaii, Mississippi, New Mexico, North Carolina and Utah. Their statutes are live, complete and queryable. What is held is the automated re-pull, so they are absent from GET /boards rather than listed as idle. A board is switched on only once its refresh has been shown to work, which makes GET /boards the list of what can be watched and never the list of what is covered. For these seven, read currency from GET /us/statutes/coverage.
Colorado is checked weekly but re-pulled annually, and that is deliberate. The Colorado Revised Statutes are published once a year with no process for periodic updates, so Colorado’s weekly slot re-pulls all 46 titles only when a new edition appears.
The cadence field on /us/statutes/coverage and /boards reports monthly for every state statute board except Arkansas, Georgia and Tennessee, which report quarterly. The 25 weekly jurisdictions above still report monthly. Read the field as the declared staleness window a board is alerted against, not as a ceiling on how often the corpus is re-pulled.
Pennsylvania’s Consolidated Statutes are complete; its older unconsolidated statutes are a separate later addition, prioritized for backfill.

​
Point-in-time and change history

Every corpus here can be asked what it said on a past date, and the answer states whether it is a published edition or a reconstruction. See Point-in-time law.

​
Detecting a change

GET /us/statutes/coverage is self-describing, so new corpora appear there as they land. To find sections that moved rather than corpora that appeared, use the changedSince filter on POST /us/statutes/search. See Coverage monitoring for both patterns, and Law change alerts to be pushed a change instead of polling for one. Corpora that are planned but not yet served, and what is deliberately out of scope, are on the Roadmap.

Corpus types

Every corpusType token, and how source and state combine with it.

Open source: Vaquill-AI/open-us-law

The ingestion pipeline behind this corpus is open source under a permissive license. Build your own copy, or contribute a source.
Last modified on September 17, 2026