Unity AI Gateway: How to connect agents to external MCPs securely

Summary

Databricks Unity AI Gateway provides a governed way to connect agents to external Model Context Protocol (MCP) servers while integrating management with Unity Catalog. The post addresses provider-specific OAuth app registration, client secrets, token refresh, permission auditing, and the lack of centralized visibility into agent access. It describes registering an external MCP server, choosing per-user OAuth or a shared principal, selecting GitHub, and letting Databricks handle OAuth registration, token exchange, and refresh; connections can then be tested in AI Playground or with DatabricksMCPClient and deployed through Agent Bricks. Unity Catalog supplies permissions and audit records, while MLflow Tracing captures requests and tool interactions, and supported providers include Glean, GitHub, Atlassian, Google Drive, and SharePoint across AWS, Azure, and GCP.

Context

Customers need agents to securely access external tools, but each provider requires separate OAuth app registration, client secrets, token refresh logic, and permission auditing. The post identifies authentication as a bottleneck and notes the lack of centralized visibility into which agents access which systems.

Approach / What changed

Register external MCP servers through Unity AI Gateway and govern them in Unity Catalog. Users can choose per-user OAuth or a shared principal, select a supported provider, and rely on Databricks to manage OAuth registration, token exchange, and refresh. Connections can be tested in AI Playground or with DatabricksMCPClient, deployed through Agent Bricks, and monitored with MLflow Tracing and Unity Catalog audit logs.

Takeaways

• External MCP servers registered through Unity Catalog receive fine-grained permissions, centralized audit records, and optional installation through Databricks Marketplace.
• Per-user OAuth lets agents act with the end user's permissions, while OAuth scopes can further limit actions such as restricting a GitHub connection to read-only repository access.
• MLflow Tracing records requests, tool calls, MCP interactions, inputs, and outputs; Unity Catalog audit logs show who accessed what, when, and through which agent.

Related reading