Use managed analytics servers
Ready-to-use endpoints cover Genie One, Genie Agent, AI Search, Databricks SQL and Unity Catalog functions.
Documentation ↗MCP server / data & analytics / source-linked
Databricks MCP
Databricks MCP is a governed tool layer for workspace analytics and retrieval. Its useful distinction is managed versus external: Databricks can host Genie, AI Search, SQL and Unity Catalog function servers, while MCP Services register external servers with grants, policies and audit logs.
Before you choose.
Identify the implementation and the job it supports.
Which implementation is this?
Databricks provides managed MCP servers and MCP Services governed through Unity Gateway and Unity Catalog; managed servers are currently public preview.
Implementation source ↗
Databricks provides managed MCP servers and MCP Services governed through Unity Gateway and Unity Catalog; managed servers are currently public preview.
Implementation source ↗
Databricks MCP is a governed tool layer for workspace analytics and retrieval. Its useful distinction is managed versus external: Databricks can host Genie, AI Search, SQL and Unity Catalog function servers, while MCP Services register external servers with grants, policies and audit logs.
Setup and connection.
The documented interface, authentication and installation path.
- Connection / transport
- Remote MCP over workspace-specific URLs
- Authentication
- Databricks on-behalf-of-user OAuth scopes such as ai-gateway, genie, ai-search, sql or unity-catalog
In the workspace, open AI Gateway → MCPs, select a managed server or register an MCP Service, then connect your agent using the workspace URL and required OAuth scope. See the client guide for agent-specific authentication.
Setup reference ↗What it actually exposes.
Concrete capabilities with maintainer sources.
Govern external MCPs
MCP Services register external servers as Unity Catalog securables with grants, policies, audit logs and usage tracking.
Documentation ↗Connect custom agents
Databricks documents discovery, authentication and tool calls from custom agent code before deployment on Databricks Apps.
Documentation ↗Evidence and method.
What was checked and what was not tested.
HOW
Source method
Public maintainer documentation and repositories. This is not a hands-on product review.
WHEN
Last checked 2026-10-01
Setup examples are documented examples. Suggested trials and comparison advice are editorial guidance.
Sources support the adjacent factual statements. A source review does not establish uptime, performance, security or current account eligibility. Read the editorial policy.
Try a specific task.
A bounded evaluation you can reproduce in your own environment.
Proposed workflow: expose governed search to an agent
Suggested evaluation, not an observed test result.
- Choose an AI Search or Genie resource and map its Unity Catalog grants.
- Register or select the managed MCP server and request only its OAuth scope.
- Connect a staging agent, inspect tool calls and audit logs before enabling production access.
What to inspect: A workspace-governed retrieval workflow. No Databricks workspace or tool call was tested.
Workflow reference ↗Permissions and limitations.
Read these constraints before granting access.
Managed MCP servers are explicitly marked Public Preview; availability and behavior can change.
Boundary reference ↗Access requires a Databricks workspace, Unity Catalog permissions and the matching OAuth scope; these are not generic public endpoints.
Boundary reference ↗Decide by workflow.
Make the expected output explicit before you connect.
A workspace-governed retrieval workflow. No Databricks workspace or tool call was tested.
Keep the task inputs, tool calls, output and any errors together when you evaluate. The trial above is a proposal, not a recorded result.
Compare the alternatives.
Specific reasons to choose another route, with the tradeoff shown.
| Alternative | Choose it when | Tradeoff |
|---|---|---|
| Snowflake MCP | Editorial choice: governed data lives in Snowflake Cortex and RBAC. | Snowflake has its own tool limits and OAuth roles. Documentation ↗ |
| MongoDB MCP | Editorial choice: an operational document database needs agent access. | MongoDB focuses on data operations rather than lakehouse governance. Documentation ↗ |
Questions before setup.
Answers specific to this implementation.
Does Databricks host every MCP server?
It hosts managed servers, and it can also register external servers through MCP Services or host custom MCPs as Databricks apps.
Documentation ↗
What does Unity Catalog add?
It enforces permissions and manages credentials so an agent reaches only the tools and data granted to it.
Documentation ↗
Source trail.
Maintainer documentation used in this record.