Databricks Genie vs Genie One vs Genie Code: what each one does

Databricks Genie is no longer one product. It is a family of 3 that share one name and one store of business context: Genie One is the simple chat and dashboard screen for business users, Genie Agents are the curated question-and-answer spaces a data team sets up (they used to be called Genie Spaces), and Genie Code is the coding agent that works in notebooks, the SQL editor and the pipelines editor.

If you remember one rule: business users ask in Genie One, data teams prepare Genie Agents so those answers are right, and engineers build with Genie Code.

The 3 Genies side by side

The last row has one catch in every column: Genie's questions run as SQL on a warehouse, and that warehouse is billed as usual. Databricks' pricing page says the free part covers Genie's model usage only.

Pick a job and see which one fits:

Try itWhich Databricks Genie do you need?

The job

·Ask how sales did last week·Find the right dashboard·Answer from company documents·Set up trusted answers for a team·Define revenue once for everyone·Build a pipeline·Fix a failing notebook

Your access

Consumer accessWorkspace or SQL access

1. Genie One

   Use this

   Business users

   Its own simple UI at /one: chat, dashboards, apps, mobile

   Free for users until 31 Jan 2027; the SQL warehouse is billed

2. Genie Agents

   Data teams who set up trusted answers

   A chat over chosen tables, with example SQL and instructions

   Free for users until 31 Jan 2027; the SQL warehouse is billed

3. Genie Code

   Engineers, analysts and data scientists

   Notebooks, SQL editor, pipelines editor, dashboards, MLflow

   150 free DBUs a user a month, then pay as you go

Chat in Genie One looks for a Genie Agent that fits the question first, then searches dashboards, queries and metric views.

Genie One: the screen for people who do not write SQL

Genie One, renamed from Databricks One, is where someone in finance or sales starts. It shows the dashboards and apps shared with them, lets them star favourites and browse by business domain, and has a chat box. They never see clusters, notebooks or catalogs.

The chat is the part that changed most. When someone asks a question, Genie One first looks for a Genie Agent that covers it, and uses that agent's tables and rules if it finds one. Only if none fits does it search dashboards, saved queries and metric views on its own. It can also read company documents once an admin connects Google Drive, SharePoint, GitHub, Glean or Atlassian. Genie One: pricing, connectors and setup covers the setup.

Genie Agents: where a data team makes the answers right

A Genie Agent is what a data team builds so that "revenue last quarter" means one thing. It has a chosen set of Unity Catalog tables, example SQL queries that show how the business asks, SQL expressions for its terms and plain-language instructions. Business users ask it questions and get back SQL, a table and a chart.

This is Text-to-SQLText-to-SQLA model turning a question in plain words into a database query, so people can ask the data directly. It is only as right as its knowledge of the tables and their meaning. with guard rails. A model writing SQL against hundreds of tables guesses at what columns mean; an agent with 5 tables, a dozen worked examples and the definition of an active customer guesses far less. That is why Genie One looks for an agent first.

Agents can also be called from your own app through the Conversation API, and connected to outside tools with MCPModel Context ProtocolAn open standard for connecting AI applications to tools and data. A system offered once as an MCP server works in any application that speaks MCP. connectors, which are in Beta.

Genie Code: the agent for people who build

Databricks calls Genie Code an autonomous agent: it plans the steps, finds the right tables, writes code, runs it, reads the output and fixes its own errors. It works in notebooks, the SQL editor, the Lakeflow Pipelines Editor, AI/BI dashboards and MLflow, and a full-page view runs several chats side by side.

2 things keep it in bounds. It can only reach data your Unity Catalog permissions allow. And by default an AI classifier approves each tool action and is meant to block actions beyond your request; you can switch to approving every action yourself. It sends your prompt, table and column names, code, sample rows and cell outputs to the model, so it needs the partner-powered AI features switched on for its agent mode.

It is the only one of the 3 that is billed now. What Genie Code does and what it costs works through the bill.

One context under all 3: Genie Ontology

The 3 products read the same map of your business, which Databricks calls Genie Ontology. Part of it is what you define on purpose in Unity Catalog, such as metric views and domains. The rest Genie infers from your dashboards, queries and agents, such as "a qualified lead counts once a demo is booked". Define a metric once and Genie One's answer and Genie Code's notebook use the same one. Genie Ontology explained shows how it ranks and checks those snippets.

The cheapest way to make every Genie more accurate is a semantic layerSemantic LayerOne shared definition of business terms and metrics, such as revenue or active customer, that reports, dashboards and AI assistants all read instead of writing their own.: write your main measures as metric views and the agents and chat answer from them instead of guessing. Our data engineering course builds one for a small shop in metric views, then puts Genie on top of it.

Which to turn on first

• Give business users Genie One once you have 2 or 3 good dashboards to show them. It costs nothing in model usage until the end of January 2027, and Consumer access keeps them out of everything else.
• Build one Genie Agent per busy subject, such as sales or support, before you announce the chat. The agent is what makes the answers trustworthy.
• Let engineers use Genie Code with a budget set first. Each user's 150 free DBUs a month reset on the 1st, and Genie budgets can block a user who passes their limit.

Genie is not the only Databricks product with a new name. Databricks renamed products lists every rename, from DLT to OpenSharing, with what changes in your code.