Build and deploy AI agents with the Agent Bricks CLI

The Agent Bricks CLI (databricks-agentbricks) is a command-line tool for developers to build custom agents directly from the terminal. It scaffolds a project from a template, runs local tests using Databricks Model Serving, and deploys to the platform's managed runtime while setting up memory, sessions, tools, and MLflow tracing automatically.

Use cases

Imagine you need to bring an AI agent to production without spending time manually wiring up infrastructure, credentials, and tools. With it, you can:

• Scaffold a new project using frameworks like LangGraph or the OpenAI Agents SDK.
• Run the agent locally on port 8000 for testing before pushing to the cloud.
• Connect managed tools via Model Context Protocol (MCP) or Unity Catalog using simple commands.
• Track execution with MLflow enabled by default.

How to use

First, install the CLI tool and log in to your workspace:

pip install databricks-agentbricks
databricks auth login --host https://<your-workspace-url> --profile <profile>
agentbricks login --profile <profile>

What happens in this code:

1. Installs the CLI Python package.
2. Authenticates to Databricks using the browser-based OAuth flow and saves the profile.
3. Sets the profile as the default for subsequent commands.

Next, scaffold a new project based on the LangGraph template and configure the stores:

agentbricks init --framework langgraph my-agent
cd my-agent
agentbricks sessions bind my-agent-sessions
agentbricks memory bind my-agent-memory

What happens in this code:

1. Creates a project directory with scaffolding and the agent.toml configuration file.
2. Navigates into the newly created project folder.
3. Binds and creates a managed session history store.
4. Binds and creates a managed long-term memory store.

To test locally and then deploy to production:

agentbricks dev
agentbricks deploy my-agent

What happens in this code:

1. Starts a local development server integrated with Databricks model serving on port 8000.
2. Provisions declared resources and rolls out the agent to the managed runtime.

Limitations and requirements

• This feature is in Beta.
• Requires the Databricks CLI installed and Python 3.10 or above.
• The migration flow for existing projects (--existing) supports LangGraph and the OpenAI Agents SDK.