Skip to content

court_dockets

Court dockets of APA suits against agencies

One row per court docket of a suit challenging federal agency action under the Administrative Procedure Act (nature of suit 899), mostly in district courts, from CourtListener (courtlistener.com: its v4 search API and its bulk docket edition) by build_courtlistener. It is the litigation counterpart to the rulemakings in dockets/documents, not all federal litigation: petitions straight to a court of appeals, such as challenges to FCC orders under the Hobbs Act, are largely absent. For court decisions, including appeals and Supreme Court cases, search court_opinion_clusters by case_name. There is no RIN, Federal Register or agency identifier on a court docket: the defendant agency appears only as text in case_name and parties_json, so a name search finds candidates, not confirmed links, and cause names the statute invoked. Primary / dedup key is cl_docket_id, a CourtListener docket object, not a case: CourtListener keeps more than one docket object for some cases (a criminal case's per-defendant dockets, or duplicate records of one filing, among them an older object reused for a new case), so one case can be several rows. Count cases on (court_id, docket_number) rather than rows, bearing in mind that a court can reuse a number for a different case, and see court_docket_groups for which rows are records of one case. All columns are stored as VARCHAR; array fields are JSON strings.

Coverage. True range with a density caveat: dockets filed from 1992 on, but only the nature-of-suit 899 (Administrative Procedure Act) selection, not the full docket record. Party names of dockets from the bulk edition are read a bounded slice per run; until read, parties_json is NULL. Recall against the publisher's nature-of-suit coding is unmeasured. (measured 2026-09-28)

  • Parquet file: court_dockets.parquet
  • MCP query_sql support: Configured; requires an available artifact.
  • Publication status: Not established by this schema page or its measurement date.
  • Row count: Not stated here; the MCP describe_table reply gives the live count under publication.
Column Type Description
cl_docket_id VARCHAR CourtListener docket id (e.g. 73613631). Primary key / dedup key of a docket object, not of a case: one case can be several rows (court_docket_groups). Distinct from the regulations.gov docket_id.
case_name VARCHAR Short case name (e.g. HENNEPIN COUNTY, MINNESOTA v. U.S. DEPARTMENT OF HEALTH AND HUMAN SERVICES). Names the defendant agency as text, not as an identifier. For decisions in cases missing here, including appeals of agency orders, search court_opinion_clusters.case_name.
case_name_full VARCHAR Full case caption when available; often blank.
court_id VARCHAR CourtListener court identifier (e.g. dcd D.D.C., cadc D.C. Circuit, akd D. Alaska).
court VARCHAR Human-readable court name (e.g. District Court, District of Columbia).
court_citation_string VARCHAR Court citation abbreviation (e.g. D.D.C.).
docket_number VARCHAR Court docket number (e.g. 1:26-cv-02460).
date_filed VARCHAR Date the case was filed (ISO 8601 string). Sort key and incremental watermark.
date_terminated VARCHAR Date the case was terminated, if closed (ISO 8601 string). Null for pending cases.
date_argued VARCHAR Date of oral argument, if any (ISO 8601 string). Often null.
nature_of_suit VARCHAR Nature-of-suit label as CourtListener spells it. Every row came from the 899 (Administrative Procedure Act) selection, but the label varies: with or without the number (appellate dockets write 2899 or 3899), abbreviated as APA or Admin Proc Act, blank or NULL, and on a few rows another code. Filtering on one exact string misses most rows; you do not need this column to find APA suits.
cause VARCHAR Statutory cause of action as recorded in PACER (e.g. 05:551 Administrative Procedure Act).
jurisdiction_type VARCHAR Basis of federal jurisdiction (e.g. U.S. Government Defendant).
jury_demand VARCHAR Jury demand recorded on the docket (e.g. None).
assigned_to VARCHAR Name of the judge assigned to the case.
referred_to VARCHAR Name of the magistrate/referral judge, if any. Often null.
parties_json VARCHAR JSON array of party names on the docket (plaintiffs and defendants, including the defendant agency). [] when the publisher names none; NULL when the docket came from the bulk edition and its names have not been read yet.
attorneys_json VARCHAR JSON array of attorney names appearing on the docket.
firms_json VARCHAR JSON array of law-firm / legal-organization names appearing on the docket.
pacer_case_id VARCHAR PACER case id for the docket in its court's CM/ECF system, as CourtListener states it. Not always the court's own number for the case: CourtListener writes 10–15-digit values on some dockets (73 rows, mostly txwd and nysd, measured 2026-10-03) and keeps an older docket object's id when it reuses that object for a new case (16 on three objects created 2020-12-22 for cases filed 2024 and 2025); some rows have none.
date_created VARCHAR Timestamp CourtListener first indexed the docket (ISO 8601 string, from the search-result meta.date_created).
absolute_url VARCHAR Absolute URL of the docket page on courtlistener.com.
case_type VARCHAR Case-type code parsed from a district-court docket_number (1:26-cv-02460 is cv, civil; also cr criminal, mj magistrate, po petty offense, mc miscellaneous). NULL when the number carries none, as appellate numbers do. The publisher codes non-civil dockets 899 too; they are kept, so filter on this for civil litigation.