Accelerate AI Development with Databricks: Discover, Govern, and Build with MCP and Agent Bricks

Data intelligence is transforming how enterprises operate, allowing organizations to build AI agents that reason over vast amounts of proprietary data. A critical component of this evolution is the ability for AI agents to connect to external services and third-party data assets to take real-world actions. This is where the Model Context Protocol (MCP) shines, acting as a standardized way for LLMs to access external tools.

However, leveraging MCP at scale presents three major enterprise challenges:

1. Discovery and Access: How can teams easily find and connect to the growing ecosystem of MCP servers?
2. Governance and Security: How can organizations enforce robust, enterprise-grade governance so that only authorized users and agents access specific servers?
3. Actionability: How can enterprises integrate these governed external services into AI agent workflows to generate insights and take action securely?

Databricks now addresses these challenges with a secure, unified, and scalable MCP experience within Agent Bricks: the MCP Catalog (Beta) for discovery and governance, an MCP in Databricks Marketplace (Public Preview) to access a growing ecosystem of external tools, and integration with the Multi-Agent Supervisor (Beta) to enable agentic orchestration.

This launch, part of Databricks’ Week of Agents, advances our vision for AI that accurately reasons on your data. By unifying governance, discovery, and real-time action across the lakehouse, Agent Bricks is making it easier than ever to build governed, production-ready agents that understand your business context and act on it securely.

Discover and Govern Tools and Agents with the MCP Catalog

We’re introducing a new ‘MCP Servers’ tab within Agents as your one-stop shop to discover, connect to, and govern all the MCP servers in your organization. Built natively on Unity Catalog, it provides the same industry-leading security, governance, and access control that enterprises already trust for their data and AI assets.

With this unified catalog, teams can:

• Discover MCP servers for tools and agents available to them across the enterprise.
• Govern access through Unity Catalog, ensuring users only connect to servers they are authorized for.
• Simplify security with built-in token exchange and end-to-end authentication.
• Audit what’s happening in your organization – not just into data and models, but also into the intelligent systems that use them.

This saves teams valuable time by reducing duplication of effort—no more building redundant tools or agents simply because other teams’ work was hard to find. It also helps enterprises minimize security risks and empower centralized teams to develop and share common MCP servers organization-wide.

The new tab also provides visibility into the managed servers already available today, including: UC Functions for custom Python or SQL execution, Vector Search for unstructured data retrieval, Genie for conversational data access, and DBSQL (new!) for running SQL queries directly against Unity Catalog tables.

We’re expanding this ecosystem with support for the Online Feature Store (Public Preview). Built on the Lakebase platform, it delivers sub-25 ms feature lookups at massive scale through a REST endpoint designed for high-QPS, agentic workloads. This integration enables agents to combine governed, real-time features with external actions—providing the domain-specific accuracy required for immediate, high-value execution.

Expand Your Ecosystem with MCP Servers in Databricks Marketplace

Discovery shouldn't stop at the servers already in your internal catalog. More and more agents are leveraging external services to take real-world actions such as requesting customer 360 summaries from company knowledge bases or retrieving real-time structured business data from trusted providers. 

The addition of MCP in Databricks Marketplace creates a central destination for discovering, evaluating, and connecting to new external MCP servers that can power your next generation of AI agents. 

Databricks Marketplace is already the trusted hub for finding and deploying AI-ready datasets, notebooks, and solutions. With the addition of MCP Servers, data and AI teams can now discover and configure external tool connections with a single click. For partners, publishing an MCP Server on Marketplace provides immediate reach to thousands of Databricks customers who can seamlessly connect Agent Bricks to their services.

The Marketplace also keeps customers current as new tools enter the ecosystem, enabling instant adoption by AI agents. Whether enriching agents with real-time financial data, coordinating with external agents, or taking actions across enterprise systems, the MCP Server catalog in Databricks Marketplace makes it simple to extend your ecosystem.

Launch partners include You.com, Glean, S&P Global, Factset, Dun & Bradstreet, Moody’s and more — giving your agents instant access to high-quality, AI-ready data and tools.

"Customers need faster, more intelligent ways to connect insight to action. We’re excited to expand our collaboration with Databricks through the launch of two MCP Servers on Databricks Marketplace, giving customers access to a range of trusted S&P Global data. Through these integrations, customers can now connect our trusted financial and commodity intelligence with their own data and AI workflows, unlocking real‑time insights, accelerating analysis, and transforming the way decisions are made across industries" — Bhavesh Dayalji, Chief AI Officer at S&P Global and CEO of Kensho

“For over a century, Moody’s data has powered critical decisions across global markets. Now, through Databricks Marketplace, the same trusted data that spans over 600 million company records, credit ratings, and proprietary research is available within Agent Bricks through access to Moody’s MCP servers. We are thrilled to empower customers to transform their business development, compliance, credit analysis, and early warning monitoring workflows through direct access to Moody’s data, insights and analytics optimized for GenAI solutions and available directly where they are innovating.” — Cristina Pieretti, Head of Digital Content & Innovation, Moody's

“With MCP servers now available in Databricks Marketplace, organizations can integrate Dun & Bradstreet’s trusted business context directly into Agent Bricks. Anchored by the D-U-N-S® Number, this integration reduces hallucinations, enhances decision accuracy, and accelerates the deployment of agentic systems.” — Anna Krayn, GM of Partnerships & Alliances at Dun & Bradstreet

"FactSet’s AI-ready data is in high demand by our clients. We are excited to launch our MCP server on Databricks Marketplace. Through collaborations across the broad array of AI capabilities available via Databricks, including Agentbricks, Databricks One, and now MCP in Databricks Marketplace, clients’ data entitlements from FactSet can be brought into their existing workflows to drive both insights and efficiencies. Tools like those available via Databricks, coupled with our open and flexible approach, empower rapid innovation as we unify and enrich data into our clients’ AI workloads." — Patrick Starling, SVP Data Solutions, Factset

Leverage Agent Bricks Multi-Agent Supervisor to take action with MCP

Agent Bricks’ Multi-Agent Supervisor (MAS) coordinates multiple agents to generate deep insights from your enterprise data — and now, with MCP integration, those agents can access third-party, AI-ready data from across your catalog or the Marketplace or trigger actions in your external systems. The result is a seamless, governed experience that enables agents not only to retrieve information, but to act on it — running analyses, filing tickets, and producing results that matter.

“With the Agent Bricks Multi-Agent Supervisor, our marketing leaders get richer insights from every campaign. It blends Genie’s data with Knowledge Assistant’s context to explain results like ROAS in a way that drives smarter decisions.”— Bruno Rocha, Digital Technology Director, Sociedade De Ensino Superior Estácio (YDUQS)

Get the Full End-to-End Experience: From Discovery to Action

Ready to start building powerful, governed agents? The new MCP features are available in your Databricks account today.

• Join us on 11/11 for “The Future of AI: Build Agents that Work” with Sam Altman and Ali Ghodsi.
• Watch the demo video
• Discover and Govern: Leverage the MCP Servers tab to explore and manage the MCP servers available to your organization.
• Build with Agent Bricks Multi-Agent Supervisor and MCP: Start integrating external actions into your agents by connecting governed MCP servers through the Multi-Agent Supervisor in Agent Bricks.
• Learn about the other announcements we’ve made in the Week of Agents