Introducing Agent Memory on Databricks
Databricks, Tuesday, October 20th, 2026
Databricks launches agent memory in beta on Agent Bricks, built on its MemCurrent research and running on Lakebase.
Databricks has introduced agent memory on Agent Bricks, its agent development platform, now in beta.
Agent memory stores key facts, preferences and workflows so agents avoid rediscovering information or repeating corrected mistakes.
It runs on Lakebase and is built on the MemCurrent architecture from Databricks Research.
Memory is available through plugins, on-demand tools and an offline dreaming pipeline that consolidates memories in the background.