docketx/us-caselaw
US Caselaw Code & tools: github.com/docketx — legal-scrambler pseudonymises a case file on your own hardware before a frontier model sees it; claude-for-legal is the Claude Code plugin (docketx-open-law) that loads these datasets and checks citations against them. The American case-law record in one dataset — published opinions from the Free Law Project / CourtListener bulk export of 2026-06-30: full text, all opinion types (majority, concurring, dissenting, per curiam… See the full description on the dataset page: https://huggingface.co/datasets/docketx/us-caselaw.
US Caselaw
Code & tools: github.com/docketx — legal-scrambler pseudonymises a case file on your own hardware before a frontier model sees it; claude-for-legal is the Claude Code plugin (docketx-open-law) that loads these datasets and checks citations against them.The American case-law record in one dataset — published opinions from the Free Law Project / CourtListener bulk export of 2026-06-30: full text, all opinion types (majority, concurring, dissenting, per curiam, orders), one uniform row format across every jurisdiction.
10,797,772 opinions are live in this repo across 50 states + DC, the Supreme Court, the federal appellate, district, bankruptcy and special courts, the U.S. territories, the tribal courts, and a catch-all of state trial courts, attorney-general opinions and specialty tribunals that no other series carries. Every number on this page was measured by counting the lines of the actual uploaded file; none is an estimate.
Load it
from datasets import load_dataset
ds = load_dataset("docketx/us-caselaw", split="train") # everything
ds = load_dataset("docketx/us-caselaw", split="train", streaming=True) # 60 GB — stream itOne jurisdiction only? Each also stands alone as docketx/us-caselaw-<code> (e.g. docketx/us-caselaw-tx), which is much cheaper than pulling the union.
Row format
Every row is docketx record v1. Every column is a string in every file:
The schema is declared in dataset_info above rather than inferred, and the opinions config lists its files explicitly. That is deliberate: a blind data/*/*.jsonl.gz glob previously pulled statute and court-rule files (which carry rule_set/rule_number and a differently-shaped extra) into the same config, so the parquet conversion inferred its schema from the first file, hit a column-name mismatch, and failed the whole dataset to 0 rows. Statutes and rules now live only in docketx/us-statutes and docketx/court-rules, where they belong.
What is here — measured counts
States and DC
Federal, territorial, tribal and catch-all
Every count on this page was measured on 2026-09-17 by decompressing each file the opinions config loads and counting its lines. Every state's loaded data is byte-identical to its standalone docketx/us-caselaw-<code> dataset (same SHA-256; Texas as three parts holding the same 662,717 rows), and each of those counts matched its publish manifest exactly. Total: 10,797,772 opinions.
The bulk export holds 10,798,347 opinions. The extras slice (added 2026-09-17, 477,273 rows, 0 duplicate ids, slice sha256 9123f2d3cdff3c48…) carries the 1,612 state courts that no per-jurisdiction plan included (477,294 opinions; 21 of them have empty text). It includes 67,524 attorney-general opinion reports from 16 states, which are official opinions but not court decisions. To exclude them, drop rows whose court is extras- followed by one of: texag, oklaag, arkag, laag, flaag, nyag, nebag, moag, wisag, calag, kanag, coloag, mdag, washag, minnag, paag. (Do not filter on a trailing ag: Ohio's Portage County courts and an Onondaga County court also end that way.) 575 opinions are in no file, each for a known reason: 533 in planned courts and 21 in extras have empty text, and 21 have a cluster or docket missing from the bulk export itself.
Superseded files still on disk
Until 2026-09-17 this page's config loaded an older, smaller slice for 44 jurisdictions (the page then served 9,095,360 of 10,320,499; Florida loaded 259,051 of 447,040, California 227,809 of 242,831). The config now loads the complete opinions-2026-06-30.v1.jsonl.gz for those 44. The older files are no longer part of any config and are kept only until they are removed in a cleanup commit.
Coverage honesty
The open record is appellate-heavy. Most US trial-court rulings are published nowhere at all — no database has them, including this one. What is here is the published record.
Provenance and license
Judicial opinions are edicts of government and carry no copyright (Banks v. Manchester, 128 U.S. 244 (1888); Georgia v. Public.Resource.Org, 590 U.S. 255 (2020)). Packaging is CC0 1.0. Source: CourtListener bulk export of 2026-06-30, Free Law Project — https://free.law.
Built by DocketX — the same corpus powers the verified legal API at docketrouter.ai.
