Free Law Project / CourtListener

CourtListener is the main platform of the Free Law Project, a nonprofit that builds open legal tools. It gathers written opinions from federal and state courts, oral argument audio, citation networks between cases, and financial disclosure filings for judges. Its RECAP Archive, assembled from federal PACER records that users have contributed, adds hundreds of millions of docket entries and many of the documents attached to them, which is the closest thing to an open window into US federal litigation.

You have several ways in. The REST API is good for targeted lookups and for pulling specific courts or date ranges, the bulk data exports let you load a large slice of case law in one go rather than scraping page by page, and there is a citation lookup tool if you mainly need to resolve references. For a retrieval corpus, the bulk files are usually the sensible starting point, with the API layered on for keeping it current.

For RAG, this is the widest open foundation for a US legal assistant: case law you can chunk and embed, plus citation links you can use to follow precedent or rank results by how often an opinion is cited. Most US court opinions are federal or state government work and therefore in the public domain, so the underlying opinion text is genuinely free to reuse, including commercially.

The caveats are practical. Coverage and metadata quality vary by court, older opinions can arrive as scanned PDFs that need optical character recognition before you can index them, and RECAP holds docket entries and filings that are not always complete for a given case. The API is rate limited, so plan bulk work around the exports. And while the opinions themselves are public domain, be careful with any third-party editorial additions, such as headnotes or summaries, that may carry their own rights.

For US legal RAG it is the natural first stop; for corporate filings you would add EDGAR, and for statutes and regulations a source such as the Code of Federal Regulations.