Legacy agent offerings

Legacy agent offerings are earlier Databricks agent products that are no longer best practice for building agents. They remain documented so that you can maintain and migrate existing agents. Don't use them for new agents. For how Databricks defines support lifecycles, see Databricks support lifecycles.

For new agents written in code, Databricks recommends the Agent Bricks CLI, which serves your agent with DurableAgentServer and deploys it to Agent Runtime. See Deploy agents on Databricks. For new agents built without code, use Genie Agents.

Legacy offerings and their replacements

Legacy offering	Use instead
Knowledge Assistant	Genie Agents for low-code agents over your documents and data.
Supervisor Agent	Genie Agents for low-code agents. To coordinate multiple agents in code, use the Agent Bricks CLI.
Supervisor API (deprecated)	Write your own agent loop and deploy it with the Agent Bricks CLI. See Deploy agents on Databricks.
Build a custom multi-agent system	The Agent Bricks CLI. To give your agent Genie Agents as tools, run agentbricks tools add genie-agent.
Run agents on Databricks Apps using the legacy agent server	The Agent Bricks CLI, with DurableAgentServer on Agent Runtime.
Create a chat agent front-end	Projects that you create with the Agent Bricks CLI include a chat UI.
Custom Agent on Model Serving	The Agent Bricks CLI. To move an existing agent, see Migrate an agent from Model Serving to Databricks Apps.
Create a custom LLM agent	AI Functions, such as ai_query.
Managed memory stores (legacy)	Managed agent memory.
AI Playground	To query models, see Query foundation and embedding models. To build and test agents, use the Agent Bricks CLI and run your agent locally with agentbricks dev. To query a deployed agent, see Query agents deployed on Databricks.

Legacy offering	Use instead
Knowledge Assistant	Genie Agents for low-code agents over your documents and data.
Supervisor Agent	Genie Agents for low-code agents. To coordinate multiple agents in code, use the Agent Bricks CLI.
Supervisor API (deprecated)	Write your own agent loop and deploy it with the Agent Bricks CLI. See Deploy agents on Databricks.
Build a custom multi-agent system	The Agent Bricks CLI. To give your agent Genie Agents as tools, run agentbricks tools add genie-agent.
Run agents on Databricks Apps using the legacy agent server	The Agent Bricks CLI, with DurableAgentServer on Agent Runtime.
Create a chat agent front-end	Projects that you create with the Agent Bricks CLI include a chat UI.
Custom Agent on Model Serving	The Agent Bricks CLI. To move an existing agent, see Migrate an agent from Model Serving to Databricks Apps.
Create a custom LLM agent	AI Functions, such as ai_query.
Managed memory stores (legacy)	Managed agent memory.
AI Playground	To query models, see Query foundation and embedding models. To build and test agents, use the Agent Bricks CLI and run your agent locally with agentbricks dev. To query a deployed agent, see Query agents deployed on Databricks.

The Self-managed agent memory (Lakebase) page is also legacy. Use managed agent memory and managed agent sessions instead.