buriedsignals/anomaly/courtlistener-docket

Query CourtListener RECAP dockets

meta.yaml is the executable contract. Use only released operations.

<!-- BEGIN GENERATED OPERATION STATUS -->

Current catalogue release status

meta.yaml is authoritative. Execute only operations listed as Released below; other sections in this playbook may document provider or unreleased adapter scope.

Released

• us/courtlistener/docket:search-dockets — Search the RECAP index and return docket metadata plus no more than three matching filing-document summaries per docket.
• us/courtlistener/docket:get-docket — Run an exact docket_id search and return the matching indexed docket record, still subject to the three-document embedding limit.
• us/courtlistener/docket:court-cases — List one court's dockets, optionally by filing-date range, from the /dockets/ database endpoint as court domain rows.
• us/courtlistener/docket:case-parties — One request per docket id to /parties/ with nested results filtered to that docket; one row per party and role.
• us/courtlistener/docket:case-counsel — One request per docket id to /attorneys/ with nested results filtered to that docket; the endpoint publishes no firm, so firm_name is empty.

<!-- END GENERATED OPERATION STATUS -->

Search or retrieve

Use search-dockets for a party, case, filing term, or documented advanced operator. Use get-docket only for a known CourtListener numeric docket ID.

catalogue query us/courtlistener/docket --operation search-dockets \
  --input '{"q":"Purdue Pharma","court":"nysb","limit":5}'
catalogue query us/courtlistener/docket --operation get-docket \
  --input '{"docket_id":16199029,"limit":1}'

The documents array contains no more than three matching RECAP filing documents. When more_documents is true, additional matching filings exist. This operation does not enumerate the docket sheet.

Court domain rows

court-cases, case-parties and case-counsel fill the court domain contract (domains/court.yaml) from the PACER database endpoints, so their rows can be written to CSV and passed to anomaly.oneshot --tables. These endpoints need a CourtListener token (anonymous requests got HTTP 401 on 2026-10-27), and an account starts at 5 requests a minute, 50 an hour and 125 a day (https://wiki.free.law/c/courtlistener/help/api/rest/v4/rest-api-v47). Each call is bounded: at most 5 docket pages, or 10 docket ids with one request each.

catalogue query us/courtlistener/docket --operation court-cases \
  --input '{"court":"txnd","filed_after":"2024-01-01","filed_before":"2024-12-31","pages":2}'
catalogue query us/courtlistener/docket --operation case-parties --input '{"docket_ids":[4214664]}'

Evidence discipline

• Preserve the exact query, selected court, result type, retrieval time, and ID.
• Treat counts over 2,000 as approximate; CourtListener documents about ±6%

error for type r and d cardinality counts.

• Open source_url and underlying filing URLs before reporting legal claims.
• Distinguish absence from the RECAP index from absence in PACER or the court.
• Do not use the data for a prohibited FCRA purpose.

Read the API reference for authentication, limits, type r, result counts, and rights. Read the query guide for search and reporting practice.

Bundled resources

• references/api-reference.md — Firecrawl-preserved primary documentation and live contract observations.
• references/query-guide.md — query choice, truncated-document handling, and reporting checklist.
• scripts/verify.py and assets/verification-cases.json — live checks for both operations.
• evals/evals.json — behavioral cases for completeness, identifiers, and legal verification.

python3 scripts/verify.py --list
python3 scripts/verify.py