Oracle Expands Agentic AI Capabilities to Dedicated Infrastructure with Ten New Select AI Features
Oracle has made a decisive move in the enterprise AI space, announcing ten new Select AI capabilities for Autonomous AI Database on Dedicated Infrastructure (ADB-D). At the center of this announcement is the Select AI Agent framework, powered by the new DBMS_CLOUD_AI_AGENT package APIs, which brings full multi-agent orchestration directly inside the database — no external middleware, no data movement, and no compromise on data sovereignty.
For organizations running Autonomous Database on Dedicated Infrastructure — a deployment model favored by regulated industries including financial services, healthcare, and government — this release represents a significant inflection point. Enterprises can now build, coordinate, and run sophisticated agentic AI workflows exactly where their data lives: inside a hardened, dedicated database environment.
The Supervisor Agent Pattern: Orchestrating Specialized AI Workers at Runtime
Perhaps the most consequential feature in this release is the introduction of the Supervisor Agent pattern for multi-agent orchestration. This architectural pattern allows a primary supervisor agent to dynamically select and coordinate specialized worker agents at runtime, based on the nature of the incoming request.
Here’s how it works in practice: when a complex query or task is submitted, the Supervisor Agent decomposes it into subtasks, identifies which worker agents are best suited to handle each piece, dispatches work sequentially, and then assembles a coordinated response. Each worker agent can be specialized — one might handle financial calculations, another might perform natural language analysis on unstructured data, and a third might execute compliance checks against regulatory rules stored in the database.
This pattern eliminates the need for external orchestration layers that have traditionally introduced latency, security concerns, and architectural complexity. The entire coordination lifecycle — from task decomposition to response assembly — happens within the database boundary.
DBMS_CLOUD_AI_AGENT: A Purpose-Built API for Enterprise Agent Development
The DBMS_CLOUD_AI_AGENT package provides a comprehensive set of APIs that Oracle professionals will find both powerful and practical. Key capabilities include:
- Tool Discovery: Agents can automatically discover and register available tools, functions, and data sources within the database environment, reducing manual configuration overhead.
- Agent Team Coordination: APIs for defining agent teams, assigning roles, and managing the relationships between supervisor and worker agents.
- Run State Checking: Built-in mechanisms for monitoring agent execution status, handling failures gracefully, and ensuring reliable task completion.
- Memory Depth Configuration: Fine-grained control over how much conversational and operational context agents retain across interactions, enabling both stateless efficiency and stateful continuity depending on the use case.
These APIs are designed to work natively with PL/SQL and SQL, meaning database developers and DBAs can build agentic workflows using skills they already possess — without needing to become Python or JavaScript experts to participate in the AI revolution.
Interoperability Standards: MCP Server and Agent2Agent (A2A) Integration
Oracle is clearly betting that the future of enterprise AI is not siloed but interoperable. This release introduces integration with two critical standards:
- Managed MCP Server: The Model Context Protocol (MCP) server integration allows in-database agents to connect with external tools and data sources through a standardized interface, enabling consistent tool-calling patterns across heterogeneous environments.
- Agent2Agent (A2A) Server: A2A integration enables Oracle’s in-database agents to communicate and collaborate with agents running on other platforms, following Google’s open A2A protocol for cross-platform agent interoperability.
Both integrations require Release Update (RU) 23.26 or later, so organizations planning to adopt these capabilities should begin their patching and upgrade planning now.
Why This Matters for Regulated Industries
The significance of delivering these capabilities on Dedicated Infrastructure cannot be overstated. ADB-D customers chose this deployment model precisely because they need physical isolation, strict network controls, and complete data sovereignty. Until now, adopting agentic AI often meant moving data to external services or deploying additional middleware — both of which conflicted with the very reasons these organizations selected dedicated infrastructure in the first place.
With Select AI Agent framework on ADB-D, enterprises can now run AI agents that reason over sensitive data, coordinate complex multi-step workflows, and integrate with external agent ecosystems — all without a single byte of data leaving their dedicated environment.
Practical Takeaway for Oracle Professionals
If you are running Autonomous Database on Dedicated Infrastructure, this release should trigger three immediate actions. First, verify your Release Update level and plan to move to RU 23.26 or later to unlock MCP and A2A capabilities. Second, explore the DBMS_CLOUD_AI_AGENT package documentation and begin identifying use cases where multi-agent orchestration could replace manual or middleware-dependent workflows. Third, engage your security and compliance teams early — the in-database nature of this framework is a compelling story for data governance, but it still requires thoughtful access control and audit configuration.
The era of agentic AI inside the Oracle Database is no longer aspirational. It’s here, it’s on dedicated infrastructure, and it’s ready for the enterprise.