Skip to content
Releases GitHub

Databricks MCP Server

Aembit supports the official Databricks MCP servers, which let AI Agent: A software workload that authenticates to systems, requests credentials, and accesses resources, either on behalf of a person or on its own. Aembit secures AI agents with the same identity-first model it uses for any workload. User-driven agents such as Claude Desktop also carry a blended identity that ties access to both the user and the agent.Learn more query Databricks data through Model Context Protocol: A standard protocol for AI agent and server interactions that defines how AI assistants communicate with external tools and data sources.Learn more(opens in new tab) tools.

This page describes how to configure Databricks as an MCP Server: A server that implements the Model Context Protocol to provide tools, resources, or data to AI agents and MCP clients.Learn more(opens in new tab) behind the Aembit MCP Identity Gateway: A component that brokers MCP traffic between MCP clients and target MCP servers, validating authorization and presenting Aembit-managed credentials on each request.Learn more.

This guide builds the Gateway-to-Server Policy: The Access Policy that authorizes the MCP Identity Gateway to access an MCP server on behalf of authenticated users—the Gateway itself is the Client Workload. Also called right-side auth; each MCP server behind the Gateway needs its own.Learn more—the second of the two Access Policy: Access Policies define, enforce, and audit access between Client and Server Workloads by cryptographically verifying workload identity and contextual factors rather than relying on static secrets.Learn more the MCP Identity Gateway requires. You create the first, the Client-to-Gateway Policy: The Access Policy that validates which MCP client can connect to the MCP Identity Gateway and authenticates users through your Identity Provider. Also called left-side auth; you create one per MCP client and Gateway combination.Learn more, during Gateway setup.

Prerequisites

Section titled “Prerequisites”

Before you begin, ensure you have the following:

Requirements and considerations

Section titled “Requirements and considerations”

Before you configure Databricks, review these requirements and behaviors specific to the Databricks MCP server.

  • Workspace-specific URL. The MCP host and OAuth Discover URL include the workspace hostname, which differs per workspace. Each workspace needs its own Server Workload in Aembit.
  • Short default token lifetime. The default access token TTL is 60 minutes and the refresh token TTL is about 7 days. Both are configurable in the App Connection settings. If the refresh token expires, the user must reauthorize.

Choose an authentication approach

Section titled “Choose an authentication approach”

Aembit supports two authentication approaches for Databricks. The tabs in the following sections stay in sync with the approach you choose.

  • User-Based: each user authenticates with their own Databricks identity, and the Gateway injects that user’s token into MCP requests. Aembit stores and refreshes each user’s tokens individually.
  • Admin-Based: an administrator completes the OAuth flow once during setup, and all users who access Databricks through the Gateway share the resulting tokens.

Create the OAuth app connection

Section titled “Create the OAuth app connection”

Databricks doesn’t support OAuth Dynamic Client Registration, so an administrator must pre-register one OAuth App Connection in the Databricks Account Console before users can authenticate. Both authentication approaches share this connection.

  1. Log in to the Databricks Account Console.
  2. Click the Settings icon in the sidebar.
  3. Open the App connections tab.
  4. Click Add connection and configure it:
    • Give the connection any name.
    • Set Redirect URLs to the Aembit Credential Provider’s Callback URL. Copy the exact read-only value from the Credential Provider after you create it in Aembit.
    • Set Access scopes to the APIs you need (for example, ALL APIs).
    • Check Generate a client secret.
    • Token TTLs are optional; the defaults are 60 minutes for access tokens and 10080 minutes (about 7 days) for refresh tokens.
  5. After saving, copy the Client ID and Client Secret from the connection details.

Configure the Credential Provider

Section titled “Configure the Credential Provider”

Create an MCP User-Based Access Token Credential Provider: Credential Providers obtain the specific access credentials—such as API keys, OAuth tokens, or temporary cloud credentials—that Client Workloads need to authenticate to Server Workloads.Learn more in Aembit.

  1. Log into your Aembit Tenant.

  2. Go to Credential Providers in the left sidebar and click + New.

  3. Configure the following fields:

    Field Value
    Name A user-friendly name
    Credential Type MCP User-Based Access Token
    MCP Server URL https://<account_id>.cloud.databricks.com/api/2.0/mcp
    Client ID The Client ID you copied earlier
    Client Secret The Client Secret you copied earlier
    Scopes all-apis offline_access
    PKCE Required On
    Lifetime 1 year

    For MCP Server URL, click Discover to populate the Authorization URL and Token URL.

  4. Click Save.

  5. Copy the read-only Callback URL from the Credential Provider.

Finish configuring the OAuth app connection

Section titled “Finish configuring the OAuth app connection”

User-Based Auth uses the shared OAuth App Connection you create in the prerequisites—both authentication approaches use the same connection.

  1. Create the MCP User-Based Access Token Credential Provider in Aembit and copy its read-only Callback URL.
  2. Confirm the App Connection’s Redirect URLs include that Callback URL (see the prerequisites).
  3. Supply the connection’s Client ID and Client Secret in the Credential Provider.

Authorize the Credential Provider

Section titled “Authorize the Credential Provider”

There’s no administrator authorization step for this credential type. Each user completes their own Databricks sign-in and consent the first time they access Databricks through the Gateway, and Aembit stores and refreshes each user’s tokens individually.

Create the Server Workload

Section titled “Create the Server Workload”

  1. Go to Server Workloads in the left sidebar and click + New.

  2. Configure the following fields:

    Field Value
    Name A user-friendly name
    Host <account_id>.cloud.databricks.com
    Application Protocol MCP
    Port 443 with TLS
    URL Path /api/2.0/mcp/sql
    Authentication method HTTP Authentication
    Authentication scheme Bearer

  3. Click Save.

Create an Access Policy

Section titled “Create an Access Policy”

This section creates the Gateway-to-Server Access Policy, which authorizes the MCP Identity Gateway to access Databricks on behalf of authenticated users.

Create an Access Policy linking the MCP Identity Gateway (as the Client Workload: Client Workloads represent software applications, scripts, or automated processes that initiate access requests to Server Workloads, operating autonomously without direct user interaction.Learn more), the Credential Provider you created, and the Server Workload for Databricks.

For step-by-step instructions, including the Client Workload settings that identify the Gateway, see Create the gateway-to-server Access Policy.

Verify

Section titled “Verify”

After authorization completes, the Aembit AI Access Authorized page lists the Databricks MCP Server as Ready. The AI agent can then call Databricks MCP tools through the Gateway.