What AI tools have MCP support?

Summary

• MCP (Model Context Protocol) is an open standard that enables AI agents and coding assistants to discover and interact with tools, data sources, and services through a standardized interface.
• Genie Agents expose two managed MCP servers: Genie Agent MCP (for querying a single agent via natural language) and Genie One MCP (for accessing Genie's analytical reasoning).
• External AI tools including Cursor, Claude Desktop, and LangGraph-based agents can connect to Genie Agents via MCP without writing SQL.
• Genie Code integrates multiple MCP servers simultaneously, including custom servers, Unity Catalog functions, and AI Search indexes.
• MCP enables vendor-neutral tool composition, allowing enterprise AI systems to combine Databricks data capabilities with external services.

What AI tools have MCP support?
MCP (Model Context Protocol) is an open standard for connecting AI agents to tools, data, and services. Instead of hard-coding integrations for each tool, MCP allows agents to discover what's available and interact through a consistent interface. This means new data sources or tools can be added without modifying the agent itself.

What is MCP and why it matters

MCP stands for Model Context Protocol. It specifies how AI agents and coding assistants should:

• Discover available tools and data sources without hard-coded lists
• Request resources (query data, invoke functions, fetch documents) through standard calls
• Handle authentication securely through token exchange
• Work across different platforms without vendor lock-in

Before MCP, agents needed custom integrations for each data source or service. With MCP, a single agent can talk to any MCP-compliant server, making it easier to connect enterprise data, internal tools, and third-party services in a composable way.

Databricks Genie MCP support

Databricks provides two managed MCP servers that expose Genie's capabilities:

MCP Server	What it does	When to use
Genie Agent MCP	Allows external agents to query a single Genie Agent by sending natural language questions. Genie writes and runs SQL, enforces Unity Catalog permissions, and returns results.	When you want an external agent (Claude, Cursor, LangGraph) to ask questions of a specific Genie Agent without writing SQL.
Genie One MCP	Exposes Genie One's reasoning and analytical capabilities to external agents. Going GA in early 2025.	When you need access to Genie's deep analytical reasoning from a third-party agent or coding tool.

Both servers authenticate via OAuth with the genie scope. The Genie Agent MCP supports up to 25 Unity Catalog tables per query and enforces all row-level security and column masking policies.

AI tools that connect to Genie via MCP

Coding Assistants

• Cursor: Code editor that can query Genie Agents while writing code, accessing live data without SQL expertise
• Claude Desktop: Claude running locally with MCP plugin support, allowing direct queries to Genie Agents

Agent Frameworks

• LangGraph-based agents built on Databricks or externally
• OpenAI Agents SDK when deployed on Databricks
• Databricks Agent Framework agents (native)

Genie Code
Genie Code integrates multiple MCP servers simultaneously, including:

• Genie Agents (query data)
• Unity Catalog functions (invoke business logic)
• AI Search indexes (vector similarity search)
• Custom MCP servers (connect your own tools)
• External MCP servers via Databricks Apps hosting

Business Applications

• Microsoft Foundry: Azure Databricks Genie can be added as a tool, enabling Foundry agents to access Genie insights

Key capabilities to evaluate

When choosing an AI tool for MCP integration, consider:

• MCP server discovery: Can the tool automatically find and configure MCP servers?
• Authentication: Does it support OAuth and secure credential handling?
• Tool composition: Can the agent use multiple MCP servers in a single request?
• Governance: Does it respect row-level security and access controls from the MCP server?
• Error handling: How does the tool handle timeouts or errors from remote servers?

FAQs

Does every AI tool support MCP?

No. MCP support is growing but not universal. Major coding assistants (Cursor, Claude Desktop) and agent frameworks (LangGraph, OpenAI Agents SDK) support it. Check your tool's documentation for MCP compatibility.

Can I use Genie Agent MCP from outside Databricks?

Yes. External agents like Claude Desktop or Cursor can connect to Genie Agent MCP endpoints hosted on your Databricks workspace. You'll need workspace authentication and the workspace hostname.

What's the difference between Genie Agent MCP and Genie One MCP?

Genie Agent MCP lets external tools query a specific configured Genie Agent (which has curated datasets and business context). Genie One MCP exposes the broader Genie One reasoning engine, enabling more general analytical queries. Genie Agent is GA; Genie One is going GA in early 2025.

Can Genie Code connect to external MCP servers?

Yes. Genie Code can be configured to connect to custom MCP servers. You can host custom servers on Databricks Apps or connect to externally managed MCP endpoints, enabling Genie Code to use tools and data beyond your Databricks workspace.