Unity AI Gateway: How to Connect Agents to External MCPs Securely
Summary
Databricks has launched Unity AI Gateway, fully integrated with Unity Catalog, to simplify the management of models, Model Context Protocol (MCP) servers, and tools for AI agents. This new offering, part of "Week of Agents," addresses the significant bottleneck of authenticating external tools like GitHub, Glean, and Atlassian. The Unity AI Gateway provides a secure, governed method for agents to access external systems, eliminating the need for teams to manage complex authentication infrastructure. It enables the registration and governance of external MCP servers via Unity Catalog, supports user-behalf access to ensure data privacy, and streamlines OAuth flows for various providers across AWS, Azure, and GCP.
Key takeaway
For AI Architects and MLOps Engineers building context-aware agents, Unity AI Gateway significantly reduces the operational burden of managing external tool authentication. You can now focus on agent development rather than auth infrastructure, leveraging Unity Catalog for centralized governance and simplified OAuth flows. This enables secure, fine-grained access to critical external data sources like GitHub and Atlassian, accelerating agent deployment and ensuring data privacy.
Key insights
Unity AI Gateway simplifies secure agent access to external tools by centralizing authentication and governance.
Principles
- Centralize external tool governance.
- Agents should act on user's behalf.
- Simplify OAuth flows server-side.
Method
Register external MCP servers in Unity Catalog, apply fine-grained permissions, and utilize managed OAuth flows for external system authentication, ensuring agents act on behalf of the end user.
In practice
- Connect agents to GitHub, Glean, Atlassian.
- Limit GitHub access to read-only repos.
- Install MCP servers from Databricks Marketplace.
Topics
- Unity AI Gateway
- AI Agents
- Model Context Protocol
- External System Connectivity
- OAuth Authentication
Best for: AI Engineer, MLOps Engineer, AI Architect
Related on AIssential
Editorial summary, takeaway, and curation by AIssential. Original article published by Databricks.