Build AI agents on Azure Databricks

An agent is an artificial intelligence (AI) system that reasons about a request with a model, calls tools to gather information or take actions, observes the results, and repeats until it can respond. On Azure Databricks, you build agents that use your governed data in Unity Catalog, and you deploy, observe, and govern them on the same platform as your data.

Choose how to build an agent

Pick the approach that matches how much code you want to write:

Approach Use for Guide
Genie Agents Agents that you configure without code, such as agents that answer questions over your tables and documents Create the Bakehouse Sales Assistant
Agent Bricks Agents that you write in code, with any framework or harness, such as LangGraph or the OpenAI Agents SDK, and any model Agent Bricks quickstart

A custom agent that you build with Agent Bricks can call Genie Agents as tools. See What is Agent Bricks?.

Build agents in code with Agent Bricks

Agent Bricks is the Azure Databricks agent developer platform. The Agent Bricks CLI scaffolds a project from a framework template, runs it locally for testing, and deploys it to the Azure Databricks agent runtime. See Agent Bricks CLI. Agent Bricks supports each stage of building a production agent:

  • Connect to models: Call models through Unity Gateway with one API, and switch models without changing your code. See Query foundation and embedding models.
  • Add context: Give your agent memory across sessions, and connect it to tools and data through Model Context Protocol (MCP) servers. See Agent memory and sessions and MCPs.
  • Deploy: Run your agent on Azure Databricks-hosted compute with an agent server and the agent runtime. See Deploy agents on Azure Databricks.
  • Observe and evaluate: Record each step that your agent takes with MLflow Tracing, and measure quality with large language model (LLM) judges, custom scorers, and expert feedback. See What is agent observability and quality?.
  • Govern: Control access to models, MCP servers, and functions, with guardrails, rate limits, and usage tracking in Unity Gateway. See AI governance guide.

To build and deploy your first agent, see Agent Bricks quickstart.

Use AI in queries and applications

To call a deployed agent or a model from a notebook, SQL, or an application, see Query agents deployed on Azure Databricks. To apply AI to data in your tables directly from SQL, such as to classify, extract, or translate text, use AI Functions. See Transform unstructured data using AI Functions.

Next steps

Resource Description
Operations and CI/CD Version control and deploy your projects, including agent code, with Declarative Automation Bundles and continuous integration and delivery (CI/CD)
Develop agents on Azure Databricks Learn how to build, deploy, observe, and govern agents with Agent Bricks