Genie One MCP reached general availability alongside file reasoning support in Genie Agents.

In September 2026, Databricks made the Genie One Model Context Protocol (MCP) server generally available [2], [19], [21]. The integration allowed external AI assistants and developer environments like Cursor to query governed metrics and business definitions grounded in Genie Ontology [20], [36]. Earlier in the month, Databricks also expanded Genie Agents with an API-accessible mode for multi-step analytical reasoning and added support for querying unstructured files stored in Unity Catalog volumes alongside tabular data [55].

To ground these agent interactions in common business definitions, Databricks introduced Unity Catalog Pages within Genie Ontology, providing a dedicated space to manage organizational terms and acronyms [22]. The platform also expanded connectivity options, allowing teams to query Genie directly within Slack, Microsoft Teams, and Google Sheets [37], [38], [39], while adding scheduled analytical tasks and document generation workflows within Genie One [30], [32], [34].

Community discussions focused heavily on evaluation, governance, and trust before rolling out spaces to end users [9], [14], [27]. Practitioners examined how the OntoRank algorithm infers definitions [26], shared techniques for building deterministic agents [47], and addressed semantic discrepancies where agent reasoning diverged from company-specific conventions, such as regional boundaries [49]. Other data teams worked through deployment patterns using Databricks Asset Bundles across environments [15], [46] and diagnosed integration hurdles involving citations and Unity Catalog scalar SQL functions [13], [28].