Databricks Adds Real-Time Data to AI Agents

Databricks is injecting real-time external intelligence into its Databricks Agentic Applications platform via a new MCP Marketplace. The move addresses a critical gap: AI agents that operate only on historical internal data struggle to make truly informed, autonomous decisions.

Static data limits an agent's ability to understand current market conditions, credit signals, or evolving economic factors. This forces manual research, creating bottlenecks and increasing risk. The Databricks Marketplace now acts as a conduit for these agents to access and process live, external data streams.

The platform integrates with providers like You.com for web intelligence and market sentiment, Moody's for institutional credit research, and Cotality for real estate and mortgage data. These aren't just passive data feeds; they are executable intelligence sources that agents can query, evaluate, and act upon within a governed framework. All access is managed through Databricks Unity Catalog, ensuring auditability and compliance.

Bridging the Knowledge Gap

By connecting agents to live data, Databricks aims to bridge the knowledge gap that hinders autonomous decision-making. This eliminates the need for manual data aggregation and analysis.

For instance, a loan approval agent can now incorporate real-time credit ratings, market trends, and property valuations alongside internal loan books. This allows for more nuanced risk assessment than relying solely on historical customer data.

Stateful Operations with Lakebase

Beyond real-time data access, Databricks is also enhancing agent memory and state management with Lakebase. This serverless database is designed to store agent context, decisions, and audit trails across multi-step workflows.

Lakebase ensures that agents can build upon insights gathered over time, even when interacting with different external intelligence sources across multiple days. Every decision is logged, providing a complete audit trail for compliance.

Commercial Loan Decision Support Example

A commercial loan decision support application illustrates this integration. An agent retrieves context from Lakebase, queries external intelligence from the Marketplace (Moody's for credit, Cotality for property, You.com for market outlook), synthesizes this with internal loan data, and then surfaces a proposed decision with reasoning via Databricks Genie. A credit officer can then review and approve with a single click, with the entire process logged for compliance in Unity Catalog.