Factory

Solution

Factory standardized on MongoDB Atlas to power its agent-native software development platform after finding that a patchwork of Firebase, PostgreSQL, and S3 could not deliver the mix of flexibility, retrieval quality, and scalability it needed. Its product processes highly unstructured data, associates text and metadata with embeddings, and runs code-retrieval workloads that must keep improving as the application evolves. Atlas gives Factory document and vector-database capabilities in one place, letting the team store variable structures and retrieve them efficiently without maintaining separate systems. Factory also relies on Voyage AI embeddings, which became core to its retrieval benchmarks and broader platform performance after outperforming other models. Together, Atlas and Voyage let the company process billions of tokens daily, support hundreds of thousands of developers, and keep the engineering team focused on product capabilities instead of database integration work.

Data flow

Source-code context, metadata, and embeddings are stored together on MongoDB Atlas, Voyage AI produces retrieval embeddings, and Factory’s Droids query that combined data to understand, review, test, and document software across large codebases.