KeewanoDB: AI's New Event Memory

Keewano has launched KeewanoDB, a new database architecture focused on preserving the complete sequence of events for AI agents. This innovative approach directly addresses the limitations of traditional databases that often discard event history, hindering AI's ability to understand causality and context. KeewanoDB stores events in their original sequence, attaching them directly to the entity they pertain to.

This initiative is supported by $12 million in seed funding led by Hetz Ventures. KeewanoDB is currently available as a fully managed service on Google Cloud, with plans for expansion to other cloud platforms and self-managed deployments by early 2027.

The database's architecture features a unique storage format that sequencers events by entity rather than across tables, eliminating the need for joins. A semantic layer, powered by large language models during ingestion, adds context to each event, with changes persisting in the event stream.

KeewanoDB's query path allows agents to run scripts directly within the database via the Model Context Protocol, streamlining data access. This contrasts with many current agent architectures that require data to be moved externally.

Industry analysts note KeewanoDB's native integration of machine learning capabilities as a key differentiator. This approach avoids the performance and resource issues often associated with retrofitted ML functionalities in traditional databases.

Research indicates that a significant percentage of enterprises struggle with AI agents producing incorrect outputs due to missing or inconsistent context. KeewanoDB aims to resolve this by providing agents with the structured event history they require to function effectively.