CourtListener API — what it is, where it stops, and the LawDiver alternative

What the CourtListener API is good at

Free Law Project runs a public research API and publishes bulk dumps that seeded much of modern open caselaw work. For journalists, academics, and exploratory scripts, CourtListener is often the right first call. LawDiver uses CourtListener dumps as a documented harvest gap-filler — see Licenses — and does not pretend Free Law Project is unimportant.

Be precise about the strengths. Bulk exports cover courts, dockets, opinions, clusters, citations, and parentheticals as CSV snapshots regenerated quarterly. The REST API exposes opinions, dockets, courts, judges, oral arguments, RECAP filings, and alerts. The citation graph works in both directions, with depth counts and parentheticals built from Free Law Project’s eyecite work. Since April 2024 a citation lookup endpoint can parse a block of text and tell you whether each case citation resolves — up to 250 citations per request. And since late 2025 CourtListener has shipped semantic search (API November 2025; website May 2026). Anyone still calling it keyword-only has not looked recently.

Hugging Face CAP dumps are not a living corpus

The Hugging Face mirrors of the Caselaw Access Project are the fastest way to get bulk American case law onto a GPU — roughly 6.7 million state and federal decisions, CC0-licensed. The problem is not subtle: Harvard’s scanning covered official book-published U.S. case law through the last volumes received in 2018. CAP handed ongoing collection to CourtListener in March 2024 and retired its own API. A product trained on the Hugging Face CAP snapshot alone is reasoning about American law as it stood roughly eight years ago — before Dobbs, Bruen, and Loper Bright. For pretraining and NLP research those datasets are excellent. As the retrieval corpus behind a filing product, a 2018 cutoff is disqualifying on its own.

Retrieval is not research

Finding the case is maybe forty percent of legal research. Everything expensive happens after you find it. Free case data ships with none of the validation layer fully in place:

• Good-law verification. CourtListener shows citing opinions and parentheticals; it does not classify treatment as followed, distinguished, questioned, or overruled. Free Law Project’s AI-assisted citator has been in progress since 2025 and, as of the comparison sources used here, has not shipped a product signal. Silence reads as approval. LawDiver ships GoodLaw status on every search result and API response, with an explicit unknown flag apart from negative so an app cannot mistake “not yet determined” for “cleared.”
• Search you can steer. CourtListener now has keyword, semantic, and hybrid modes. LawDiver exposes citation, case-name, boolean-with-proximity, and semantic engines plus selectable modes and an auto router, so a mixed query can run more than one engine at once and jurisdiction phrases become filters rather than search terms.
• Authorities that are not cases. CourtListener’s graph is strong on case-to-case citation. Real briefs also follow statutes, regulations, and court rules. LawDiver’s extraction classifies authorities by kind so the graph includes those sources.
• Document-level cite checking. CourtListener’s lookup answers whether a citation string resolves to a real case. It does not compare caption to reporter cite, check year or court, correct Bluebook form, or say whether the authority is still good law. LawDiver’s CiteCheck and cite-check API cover that workflow.
• Currency in minutes, not quarters. If the last export ran on June 30 and the overruling opinion came down on July 12, a dump-backed product is confidently wrong for up to three months — and will not know it. LawDiver harvests continuously from court sites, feeds, and government publications.

Where the CourtListener API is inferior to LawDiver for products

Inferior here is not an insult to a nonprofit — it is a product comparison. If you are shipping software people pay for or depend on in a filing workflow, CourtListener’s defaults lose to LawDiver on the dimensions that decide launch:

• Throughput. 5/min and 125/day cannot support cite-check-as-you-type, parallel fan-out, or a nightly sync. LawDiver defaults to a 60/min-class ceiling with a much larger rolling day budget, raised per account on request.
• Commercial licensing transparency. CourtListener’s membership API is not a blanket commercial license. LawDiver publishes any-commercial-use terms on the API product page.
• Good-law and multi-authority cite-check. CourtListener citation lookup is primarily existence/resolution for cases, with its own batch caps. LawDiver returns treatment signals and cite-checks statutes, C.F.R., rules, and specialty authorities on the same key.
• Live corpus vs. quarterly bulk. Product UIs need a continuously harvested store. Dump-based pipelines are real work; LawDiver already runs that work behind the API.

As a data source, CourtListener remains the right starting point for research, journalism, and nonprofit work. As a finished filing product, free data alone is not enough — that gap is where KeyCite, Shepard’s, and LawDiver GoodLaw live. Deep dive on the throttle numbers: CourtListener API rate limits.

Need the rate-limit deep dive?

Numbers, 429 behavior, citation-lookup caps, and window mechanics are documented on Free Law Project’s own REST docs. LawDiver’s detailed comparison of those throttles lives for teams who already know they are blocked by 429s — this page is the broader CourtListener API overview for builders choosing a stack.

Attribution

CourtListener is a service of Free Law Project and does not sponsor or endorse LawDiver. Rate-limit figures above follow CourtListener’s public REST documentation as of October 2026; if Free Law Project changes defaults, their docs control.