Azure Databricks Managed MCP Servers: Connect Your AI Agents to Data Without the Plumbing

TL;DR Azure Databricks Managed MCP servers are now in Public Preview, allowing your AI agents to securely connect to Databricks resources and external APIs . This removes a major friction point for .NET and Azure developers building production agentic workflows—no more hand-rolling authentication, token management, or proxy layers.

The Problem You’ve Been Solving Manually

If you’ve built an AI agent in the last year, you’ve probably hit this wall: your agent needs to query data in Unity Catalog, call a Databricks vector search index, or invoke a custom function. Sounds simple. Then you realize you need to:

Manage OAuth flows and token refresh yourself
Build a secure proxy to hide credentials from the agent
Handle authentication errors and retry logic
Ensure your agent can’t accidentally leak tokens in logs Anthropic’s Model Context Protocol (MCP) is a “USB-C for AI” that lets AI agents talk to external tools like databases, search engines, and APIs, and OpenAI and Microsoft have publicly embraced MCP . But until now, the integration work was still on you.

What’s New: Databricks Managed MCP Servers (Public Preview)

Databricks managed MCP servers are ready-to-use servers that connect your AI agents to data stored in Unity Catalog, Databricks Vector Search indexes, Genie spaces, and custom functions . More importantly: Databricks handles all OAuth flows and token refresh automatically, tokens are never exposed to end users, and authentication uses consistent patterns through Unity Catalog connections .

Translation: you get secure, authenticated agent tooling out of the box.

Practical Integration for .NET + Azure Developers

Setting Up Locally (Python Example First)

Ensure you have a local environment with Python 3.12 or above, then install dependencies with: pip install -U "mcp>=1.9" "databricks-sdk[openai]" "mlflow>=3.1.0" "databricks-agents>=1.0.0" "databricks-mcp"

from databricks_mcp import DatabricksMCPClient
from databricks.sdk import WorkspaceClient

workspace_client = WorkspaceClient(profile="DEFAULT")
host = workspace_client.config.host

Use OAuth to authenticate to your workspace by running databricks auth login --host https://<your-workspace-hostname> .

What This Means for .NET Developers

With .NET, you can use artificial intelligence (AI) to automate and accomplish complex tasks in your applications using the tools, platforms, and services that are familiar to you . The MCP integration opens a direct bridge: for new applications that require agentic capabilities, multi-agent orchestration, or enterprise-grade observability and security, the recommended framework is Microsoft Agent Framework .

You can now wire Databricks data sources into your .NET agents via the MCP protocol without writing custom authentication middleware. You can connect to the server using standard SDKs, such as the MCP Python SDK, and Databricks MCP servers are secure by default and require clients to specify authentication, with the databricks-mcp Python library simplifying authentication in custom agent code .

Testing in AI Playground (No Code Required)

Test MCP servers directly in AI Playground without writing any code: go to AI Playground in your Databricks workspace, choose a model with the Tools enabled label, click Tools > + Add tool and select MCP Servers, then select External MCP servers to browse available connections and choose the Unity Catalog connection you installed earlier .

Cost & Performance Considerations

Managed MCP server pricing depends on the type of feature: Unity Catalog functions use serverless general compute pricing, Genie spaces use serverless SQL compute pricing, and DBSQL servers use Databricks SQL pricing . This means you pay only for what your agents actually call—no idle proxy overhead.

Why This Matters Right Now

If 2025 was the year AI got a vibe check, 2026 will be the year the tech gets practical, with the focus shifting away from building ever-larger language models and toward the harder work of making AI usable, which involves deploying smaller models where they fit, embedding intelligence into physical devices, and designing systems that integrate cleanly into human workflows .

Managed MCP servers are that “clean integration” for data-driven agents. They let you focus on agent logic instead of infrastructure plumbing—exactly what enterprises need as they move from proof-of-concept to production.

Next Steps

Enable the preview in your Azure Databricks workspace (check your Previews page).
Test in AI Playground with an existing Genie space or Vector Search index.
Build a .NET agent using Microsoft Agent Framework and wire it to your Databricks MCP server via standard OAuth.
Monitor costs via serverless compute billing—it’s per-call, not per-hour.

The friction is gone. Ship faster.