Databricks Launches Genie One MCP to Ground AI Agents

Databricks has announced the general availability of its Genie One Model Context Protocol (MCP) server, now offered as a managed service within Unity Gateway. This release aims to solve the problem of agent sprawl, where isolated AI assistants operate on stale or conflicting business definitions. By exposing Genie One over the open MCP standard, Databricks allows any third-party AI agent to communicate with its data platform as a peer, ensuring consistent answers across an organization.

The integration relies on the Genie Ontology to provide a unified semantic layer across both structured relational tables and unstructured document repositories. Instead of building custom connectors for every data format, developers can use the Genie One MCP to give agents a single interface. The system exposes tools for asking questions, retrieving query results, tracking incremental progress, and steering responses. Additionally, the associated MCP App enables supported clients to embed real-time interactive visualizations and citations directly into their workflows.

For practitioners, this means popular tools like Claude, ChatGPT, and Cursor can now tap into a single, governed enterprise context. For example, a presentation-generating agent can query the MCP to pull verified business results directly into slide decks. Similarly, customer success agents can retrieve telemetry signals to customize outreach, while software developers can use coding assistants to fetch active product definitions during pull requests. This centralized governance model includes fine-grained policies and audit logging managed through Unity Gateway, ensuring security remains intact regardless of which external AI tool a team chooses to deploy.