Databricks brings Genie plugins to ChatGPT and Claude, so agents read governed definitions before they query business data

The frontier agents teams already use can now call Databricks enterprise context through native plugins instead of each wandering through organizational data on its own. The prior pattern was a live search in which the agent hopped between MCP servers, opened a document, read a schema, tried a table and backed out, with little context on where the right answer lived. The plugins work natively today in ChatGPT, Claude, Microsoft Copilot Cowork and Grok Bot, so teams keep business context in Databricks and make it available across those agents.

Genie turns a business question into a fixed loop that runs from term meaning to a cited answer. Genie Ontology is the governed context layer in Databricks, supplying definitions, relationships and source context so Genie can interpret business terms such as the company's agreed activation metric. Genie works out what the business terms mean, finds data the user is allowed to see, writes and runs SQL, routes to a domain Genie Agent when further investigation is needed, and returns an answer with citations; in clients supporting the MCP Apps extension, charts and Genie Ontology citations appear directly in the conversation.

The material describes native availability of the Databricks Genie plugins in ChatGPT, Claude, Microsoft Copilot Cowork and Grok Bot, plus access through the Genie MCP server for agents without a native plugin; the scenarios given are a product team tracking feature adoption, a business analyst explaining a regional revenue miss, an account team preparing meeting materials and a growth marketing agent running campaign reviews.

The material gives no accuracy, latency or comparison-against-direct-retrieval data on real enterprise data, and the example tasks are illustrative scenarios rather than evaluation results.