Crystallography Open Database
The Crystallography Open Database (COD) gathers crystal structures, the atomic-level arrangements that describe exactly how a solid is built, for over 350,000 compounds. Coverage spans organic, inorganic, metal-organic, and mineral materials, drawn from published literature and community deposits. Each entry records the unit cell, atomic positions, symmetry, and the bibliographic reference for the paper it came from.
The data comes as CIF files (Crystallographic Information File, the standard plain-text format for structural data). You can download the whole database as a single archive, sync it over rsync to keep a local mirror current, or query it through a REST interface and a SPARQL endpoint if you want to pull specific structures by formula, elements, or cell parameters. Because CIF is text with a well-defined key-value structure, it parses cleanly, and there are mature libraries (for example in Python) that read it into structured objects.
For a RAG (Retrieval-Augmented Generation) system, COD suits materials science, chemistry, and solid-state physics assistants that need to ground answers in real, referenced structures rather than guess at them. A common pattern is to turn each CIF into a short natural-language summary (composition, space group, key distances, and the source citation) and embed that, keeping the raw file linked for anyone who needs the exact coordinates. The database is already cleaned and curated, so it is closer to retrieval-ready than most raw scientific dumps.
On licensing there are no gotchas: every record is released under CC0, placing it in the public domain, so you can use the whole database in commercial products with no attribution requirement and no conditions to track. That makes COD the natural open counterpart to subscription structure databases like the CSD or ICSD when your budget or your redistribution needs rule those out.
Related sources
Bio2RDF
An open-source project that pulls together a diverse set of life-sciences datasets from many providers into a single linked-data graph, with a SPARQL endpoint for querying across them. The full collection is about 11 billion triples across 35 datasets, including DrugBank, PubMed, and MeSH.
ChEBI
A dictionary of small chemical compound molecular entities, with structure files and ontology files available for download. Particularly useful as a controlled vocabulary for chemical entity linking, where you match a mention in text to a standard identifier.
ChEMBL
A manually curated database of bioactive molecules with drug-like properties, bringing together chemical, bioactivity, and genomic data to support drug discovery. Holds close to 2.5M compound records on nearly 2M unique chemical structures, extracted mainly from the medicinal chemistry literature.
LeMaterial (LeMat-Bulk)
A harmonised, standardised merge of three major computational materials databases (Materials Project, Alexandria and OQMD) into a single format of roughly 6.7 million entries with consistent property definitions.