Startup Keewano launches agent-focused database and pulls in $12M in funding

Startup Keewano has unveiled KeewanoDB, an innovative event-oriented database engineered to empower artificial intelligence agents with real-time context for analytics and decision-making. The Tel Aviv-based company, formally known as Sandstorm Ltd., raised $12 million in funding from investors including Hetz Ventures and Andreessen Horowitz’s a16z Speedrun.

Conventional relational databases and data warehouses were originally designed for structured queries by humans, not for artificial intelligence agents analyzing long sequences of behavior. KeewanoDB groups events around an entity such as a customer, device, or transaction, storing them in chronological order. This approach allows agents to examine an entity's history without needing to flatten the information into tables or wait for an extract, transform, load (ETL) pipeline.

Each event weighs approximately four bytes and is organized on solid-state drives for efficient reading. The database operates using a distributed, multi-sharded design in Docker containers orchestrated by Kubernetes, leveraging vectorized instructions on standard central processing units rather than graphics processing units.

KeewanoDB supports Parquet and Apache Iceberg formats and can ingest event streams from systems like Apache Kafka. Customers can direct the same events to KeewanoDB for AI agents or to analytical databases like Snowflake for conventional business intelligence. The company charges based on the number of active entities rather than individual events, potentially saving customers up to 84% compared with querying their existing systems directly.

The database also features an in-database acceleration engine that executes Lua scripts sent by agents through the Model Context Protocol, allowing processing and filtering to occur within the database. The company provides a semantic engine that automatically constructs a business layer from incoming events, though customers can add manual mappings and business context.

KeewanoDB's analytics layer includes an agent called Signal, which applies machine learning and statistical methods to test hypotheses and present supporting event sequences as evidence. While the database alone does not distinguish causation from correlation, developers must specify these parameters when building agents. Keewano's separate analytics layer supports product analytics and real-time operational decisions.