Tel Aviv-based startup Keewano, formally known as Sandstorm Ltd., has introduced KeewanoDB, an event-oriented database built to provide AI agents with real-time context. The company also secured $12 million in funding from investors including Hetz Ventures and Andreessen Horowitz's a16z Speedrun.
The database organizes events chronologically around entities like customers or transactions, eliminating the need for traditional ETL pipelines. Each event occupies approximately four bytes on solid-state drives. The system uses a distributed, multi-sharded architecture running in Docker containers orchestrated by Kubernetes.
Keewano claims the system can query 250 million events in under 0.5 seconds using standard CPUs. It supports Parquet and Apache Iceberg formats and integrates with data streams from Apache Kafka. Customers can also ingest data from platforms like Snowflake and Google BigQuery.
The company charges based on the number of active entities. According to Keewano, users migrating from Snowflake or BigQuery have reported token savings of approximately 84%. An in-database acceleration engine executes Lua scripts to filter data locally, providing agents with relevant context.
The startup is launching two products: KeewanoDB for agentic analytics and a separate analytics layer for operational decision-making. This layer includes an agent called Signal, which utilizes machine learning and statistical methods to test hypotheses and identify event sequences.
