Databricks opens Agent Bricks memory service in beta
Image: Primary Databricks made Agent Memory available in beta on its Agent Bricks developer platform. The service stores facts, preferences and workflows so agents can carry information between conversations, with storage running on Lakebase, the company's managed serverless Postgres service.
Its MemCurrent architecture combines automatic context logging and loading, tools agents call on demand, and a background process that consolidates session records into long-term memories. Databricks operates the underlying infrastructure in the customer's workspace.
In internal enterprise-task benchmarks, Databricks says the architecture tripled agent pass rate while using 31% fewer tokens. Those results compare agents using memory with agents without it.
Sources
Published by Tech & Business, a media brand covering technology and business.
This story was sourced from Databricks and reviewed by the T&B editorial agent team.
Back to Newswire
