Introducing Agent Memory on Databricks

Most AI agents start every conversation from zero. Agent memory on Agent Bricks (Beta) gives your agents durable, long-term memory on Lakebase, in your Databricks workspace, so each conversation starts with what the agent learned from the last one. In this video, you'll learn how agent memory works and how to add it to your agent. Agent memory has three pieces: • Managed sessions: record each conversation, turn by turn • Managed memory: keep what's worth knowing about each user, such as preferences, facts, and workflows • Memory pipelines: run in the background to distill sessions into memory, merge duplicates, and keep memory files easy to access In Databricks research, agents with memory passed three times as many tasks (22.3% → 68.4%) while using 31% fewer tokens. Hook it up with the Agent Bricks CLI: agentbricks sessions bind agentbricks memory bind agentbricks memory pipeline create ⏰ TIMESTAMPS: 00:00 – Agents that start from zero 00:09 – Sessions, memory, and pipelines 00:36 – Results: 3x the tasks, 31% fewer tokens 00:46 – Set it up with the Agent Bricks CLI 01:03 – The next conversation remembers 🔗 LEARN MORE Add memory to an existing agent: https://docs.databricks.com/aws/en/agents/agent-memory/existing-agent-tutorial Managed memory: https://docs.databricks.com/aws/en/agents/agent-memory/managed-memory Managed sessions: https://docs.databricks.com/aws/en/agents/agent-memory/managed-sessions Memory pipelines: https://docs.databricks.com/aws/en/agents/agent-memory/memory-pipelines Agent Bricks CLI: https://docs.databricks.com/aws/en/agents/custom-agents/agent-bricks-cli Agent Bricks: https://www.databricks.com/product/artificial-intelligence/agent-bricks ❓ FAQ What is agent memory on Agent Bricks? Agent memory gives AI agents durable, long-term memory stored in Lakebase in your Databricks workspace. It has three pieces: managed sessions, managed memory, and memory pipelines. It's in Beta. What's the difference between managed sessions and managed memory? Managed sessions record each conversation turn by turn, as a full conversation history. Managed memory keeps what's worth knowing about each user, such as preferences, facts, and workflows. What do memory pipelines do? Memory pipelines run in the background between a session store and a memory store. They distill sessions into memory, merge duplicates, and keep memory files easy for the agent to access. How do I add memory to my agent? With the Agent Bricks CLI: bind a session store and a memory store (agentbricks sessions bind, agentbricks memory bind), then create a memory pipeline between them (agentbricks memory pipeline create). Does memory make agents better? In Databricks research, agents with memory passed three times as many tasks while using 31% fewer tokens. 👉 Add memory to your agent: https://docs.databricks.com/aws/en/agents/agent-memory/existing-agent-tutorial