Skip to content
Blog

Documentation

Welcome to the Kuzu docs!

Kuzu is an embedded graph database built for query speed and scalability. It is optimized for handling complex join-heavy analytical workloads on very large graphs and has the following core features:

  • Property Graph data model and Cypher query language
  • Embedded (in-process) integration with applications
  • Columnar disk-based storage
  • Columnar and compressed sparse row-based (CSR) adjacency list and join indices
  • Vectorized and factorized query processing
  • Novel and efficient join algorithms
  • Multi-core query parallelism
  • Serializable ACID transactions

If you’re new to Kuzu, or graph databases in general, we recommend starting with this introductory video. Once you’re ready to dive in, check out the rest of the documentation.

Why Kuzu?

Although there are many graph database management systems (GDBMSs) in the market today, Kuzu stands apart because its design and implementation address specific trade-offs that make it a compelling choice for analytical query workloads on large graphs. Below, we list some of the key reasons why you should consider using Kuzu.

Performance and scalability

Kuzu is a state-of-the-art graph DBMS that came out of state-of-the-art academic research and is built by a core team of database experts who relentlessly focus on scalability and performance. Its founding engineers have co-authored several cutting-edge technical papers and articles on storing and querying large graphs.

Usability

Kuzu is built for industry use cases. It implements a suite of features that lower the barrier of entry for modeling your data as a graph and querying it in an expressive graph query language. Because Kuzu is an embedded database, it runs within your application process, making it easy to set up and use Kuzu. Kuzu does not require installing any external dependencies or managing it as a DBMS server.

Interoperability

Kuzu is designed to be highly interoperable with a variety of external formats and columnar or relational stores, including Parquet, Arrow, DuckDB, and more. This allows you to easily move your existing data to and from Kuzu, making it a great choice for graph data science, machine learning, and analytics use cases.

Structured property graph model

The data model in Kuzu is based on the property graph model, together with some structure (including node and relationship tables, and a pre-defined schema). This makes it flexible and intuitive to model your existing data as a graph, while also being smart enough to optimize query performance and perform vectorized operations at scale.

Open source

Kuzu is open source and has a permissive MIT license, which allows you to build commercial and proprietary applications on top of it. Check out our GitHub repo and try out Kuzu. And while you’re there, consider giving us a star and spreading the word!