OpenAlex

OpenAlex maps the research world as a set of linked entities: works (papers, books, datasets), the authors who wrote them, the venues that published them, the institutions behind them, and the topics they cover. Each entity has an ID and points to the others, so you can walk from a paper to its authors, their institutions, and the works it cites or is cited by. It runs to hundreds of millions of works and is refreshed monthly.

OurResearch, the nonprofit behind it, built OpenAlex to replace Microsoft Academic Graph after Microsoft retired that in 2021, and released the metadata under CC0, which places it in the public domain with no reuse conditions at all. That makes it unusually easy to build on.

There are two front doors. The REST API needs no key and is ideal for lookups and for enriching records on the fly; adding your email to requests puts you in the faster "polite pool". For bulk work, download the full monthly snapshot from Amazon S3 and load it locally rather than paging through millions of API calls.

For RAG, OpenAlex shines as a metadata and citation layer around a corpus: use it to filter by field or date, rank retrieved passages by citation weight, deduplicate papers, or follow references to pull in related work. One practical wrinkle to plan for is that abstracts are stored as an inverted index (a word-to-positions map) rather than plain text, so you reconstruct the abstract before you chunk or embed it.

The important limitation is that OpenAlex gives you metadata and abstracts, not full paper text. For the documents themselves you pair it with a full-text source such as arXiv, PubMed Central, or S2ORC, and note that while the metadata is CC0, those underlying works keep their own separate licences. Used that way, OpenAlex is the connective tissue that makes a scholarly RAG corpus searchable and rankable.