Materialize Live Data Layer

Materialize, Inc. was founded in January 2019 by Arjun Narayan and Frank McSherry and keeps its permanent office in New York City, with a distributed team. It builds a SQL engine that keeps query results continuously up to date, so applications, microservices and AI agents can work on fresh operational data without overloading source databases.

Materialize ingests changes continuously from PostgreSQL, MySQL, SQL Server, Kafka and business systems, and uses incremental computation to update views and indexes when only part of the data changes. Teams define live business objects in standard SQL, link them into a context graph and expose them through a PostgreSQL interface or an MCP server for agents, while keeping vector embeddings and search indexes fresh. It is offered as a managed cloud, a self-managed edition for public or private clouds, and a local emulator. Customers named on the site include Notion, Mercari, General Mills, Ryder, Neo Financial and Zepz.

• Incremental views: Recomputes only what changed and serves results with millisecond reads.
• Change data capture: Continuous ingestion from PostgreSQL, MySQL, SQL Server and Kafka.
• Live data products: SQL-defined business objects linked into a context graph.
• Agent access: Built-in MCP server with role-based control over what each agent can query.
• Fresh search indexes: Keeps vector embeddings and indexes in Elastic, OpenSearch or turbopuffer current.
• Deployment as code: mz-deploy CLI, managed cloud, self-managed and a local emulator.

A Vietnamese digital lender can compute live loan eligibility and risk features from core banking changes. An e-commerce platform can give its customer service AI agent an up-to-date view of orders, payments and shipments. A logistics company can maintain real-time fleet and delivery dashboards without adding load to its transactional database.

Complex queries are offloaded from operational databases while results stay fresh within about a second, with single-digit millisecond reads. Customer stories on the site report an 80% cut in online feature store costs at Neo Financial, rule calculations falling from 27 seconds to 0.5 seconds at Vontive and a 30x shorter context engineering cycle at Zepz. Standard SQL replaces custom streaming code, and a free community license allows self-hosting at small scale.

AI agents and real-time applications need a consistent, current view of operational data, but querying production databases directly is slow and risky while warehouses lag by hours. Incremental view maintenance with SQL and MCP access is emerging as a way to give agents trusted context, a need that fintech, e-commerce and logistics firms in Southeast Asia share as they deploy AI assistants.

Founded in January 2019 by Arjun Narayan and Frank McSherry, with its permanent office in New York City. Raised more than 100 million USD across Series A (Lightspeed), Series B (Kleiner Perkins) and Series C (Redpoint). A free Self-Managed Community License covers installations up to 24 GiB memory.