Graphulo is a Java library for the Apache Accumulo database delivering server-side sparse matrix math primitives that enable higher-level graph algorithms and analytics. Graph primitives loosely follow the GraphBLAS spec. Example algorithms include Breadth First Search, finding a k-Truss subgraph, computing Jaccard coefficients, transforming by TF-IDF, and performing non-negative matrix factorization. We encourage developers to use Graphulo’s primitives to build more algorithms and applications. Graphulo’s design resembles that of a stored procedure in classic relational databases. The client calls (directly or through delagate functions) the OneTable or TwoTable core functions, which create new tables to store results in Accumulo rather than gather results at the client. Both functions scan Accumulo with iterators that themselves open Scanners and BatchWriters, allowing reading from multiple tables and writing to multiple tables in one client call, as opposed to ordinary scans that read fconfrom a single table and send results back to the client.