Model Context Protocol (MCP) on Databricks

MCP is an open source standard that connects AI agents to tools, resources, prompts, and other contextual information. On Databricks, MCP servers are governed through Unity AI Gateway, the enterprise control plane that governs access and monitors activity across MCP servers and LLM endpoints.

• Access control: Managed and external MCP servers use Unity Catalog permissions to control which users and service principals can access each server and its underlying data.
• Credential management: External MCP servers use Unity Catalog connections with managed OAuth to securely handle authentication without exposing credentials to end users.
• Centralized visibility: View and manage all MCP servers in your workspace from Unity AI Gateway.

Set up an MCP server

• • Install an external MCP
  • Securely connect to MCP servers hosted outside of Databricks using managed connections.
• • Host a custom MCP
  • Host a custom MCP server as a Databricks App.

Connect agents to MCP servers

Pick the type of MCP server you want to call from your agent:

• • Databricks managed MCP servers
  • Pre-configured Databricks servers for Vector Search, Genie Spaces, Databricks SQL, and Unity Catalog functions. No setup required.
• • External MCP servers
  • Call third-party MCP servers from agent code through Databricks-managed proxies.
• • Custom MCP servers
  • Connect to a custom MCP server hosted as a Databricks App.
• • Connect clients to Databricks MCPs
  • Wire up Claude, Cursor, MCP Inspector, and other clients to your Databricks MCPs.

To see your available MCP servers, go to your workspace > AI Gateway > MCPs:

Compute pricing

Custom MCP servers are subject to Databricks Apps pricing.

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.

• Databricks SQL servers use Databricks SQL pricing.

• Vector Search indices use Vector Search pricing.