Oracle has launched AI Database 26ai, its latest long-term support release that shifts focus from pure generative AI capabilities to comprehensive agentic automation. The October 2025 release introduces Select AI Agent framework, AI Private Agent Factory, and native MCP server integration—tools designed to help enterprises build autonomous workflows directly within their database infrastructure.
The new version replaces Database 23ai (originally released as Database 23c in 2023) and marks a fundamental architectural shift. Rather than treating AI as an add-on feature, Oracle is positioning intelligent agents as core database components with first-class citizenship alongside traditional database objects.
Select AI Agent Framework Targets Developer Workflows
The centerpiece of the 26ai release is Select AI Agent, a pro-code framework that extends the natural language query capabilities Oracle introduced with Select AI in 2023. Unlike conversational interfaces that simply translate questions into SQL, this framework provides developers with tools to construct, deploy, and manage autonomous agents that can execute complex multi-step operations.
Select AI Agent comes with pre-built database tools and supports external integrations through REST APIs and Model Context Protocol (MCP) servers. This architecture allows agents to combine internal database functions with external services, creating workflows that span multiple systems while maintaining centralized control within the database layer.
“This is more of a pro-code experience where developers can more specifically target database functions and knowledge context integration,” explained Alexander Wurm, principal analyst at Nucleus Research. The framework essentially transforms the database from a passive data store into an active automation platform.
Tony Baer, chief analyst at dbInsight, emphasized the significance: “It’s about making AI Agents first-class citizens inside the database.” This architectural decision means agents can access database primitives directly without the overhead of external orchestration layers—potentially reducing latency and simplifying security boundaries.
AI Private Agent Factory Extends Capabilities to Non-Technical Teams
For organizations seeking no-code deployment options, Oracle added AI Private Agent Factory alongside the developer-focused Select AI Agent. This framework targets platform and security teams who need to deploy containerized agents in controlled environments without writing code.
The Factory’s scope is intentionally narrow—it focuses specifically on deploying private agents with strict security boundaries rather than providing general-purpose agent building tools. Stephanie Walter, practice leader of AI stack at HyperFRAME Research, noted that comparable offerings include “Snowflake managed MCP plus partner kits or Google ADK patterns when self-hosting agents.”
The distinction between Select AI Agent and Private Agent Factory reflects Oracle’s attempt to serve both developer-driven automation projects and IT operations teams managing agent deployments at scale. Organizations can use the frameworks independently or combine them depending on whether they’re building custom agent logic or deploying standardized agent containers.
MCP Server Support Enables Cross-System Agent Interactions
Oracle’s integration with Model Context Protocol servers addresses a critical challenge in agentic systems: enabling agents to interact with database content while maintaining security and governance controls. The MCP server support allows developers to build agents or workflows that can query and manipulate data through standardized interfaces.
This capability becomes particularly valuable for enterprises running hybrid architectures where agents need to coordinate actions across cloud services, on-premises databases, and third-party APIs. By supporting MCP—an emerging standard for agent-to-system communication—Oracle is betting on interoperability rather than proprietary protocols.
The MCP integration works bidirectionally: external agents can access Oracle database resources through MCP servers, while Select AI Agents can call external MCP-enabled services. This creates possibilities for complex orchestrations where database-hosted agents coordinate with agents running in other environments.
Core Database Features Carry Forward with Security Enhancements
AI Database 26ai inherits key capabilities from its predecessor, including vector search for semantic queries, data annotations for metadata management, and Exadata platform support for high-performance workloads. The database also integrates with Nvidia’s NeMo Retriever microservices, enabling retrieval-augmented generation (RAG) patterns directly within database queries.
On the security front, Oracle added three layers of protection. SQL Firewall blocks unauthorized SQL activity and injection attacks at the database perimeter. Zero Data Loss Cloud Protect safeguards on-premises Oracle databases from data loss and ransomware using the Zero Data Loss Recovery Service’s continuous protection model. The database also implements quantum-resistant encryption algorithms for data in motion and at rest—a forward-looking measure as quantum computing advances.
These security features matter particularly for agentic use cases, where autonomous systems need to execute operations without constant human oversight. The SQL Firewall can restrict agent actions to pre-approved query patterns, while the quantum encryption prepares for future threats.
Simplified Migration Path and Lakehouse Foundation
Oracle is making the transition from Database 23ai straightforward: existing customers can upgrade to 26ai by applying the October 2025 update without full database upgrades or application re-certification. This approach reduces migration friction for enterprises running production workloads on the previous version.
The 26ai release also serves as the foundation for Oracle’s new Autonomous AI Lakehouse, which supports Apache Iceberg table format for cross-platform interoperability. This integration suggests Oracle is positioning the database not just as a transactional system but as a unified data platform that can serve both operational and analytical workloads with embedded intelligence.
What’s particularly interesting about Oracle’s approach is the architectural bet that the database should be the control plane for enterprise automation, not just a data repository. By embedding agent frameworks directly in the database layer, Oracle is challenging the emerging pattern of orchestration platforms that sit above databases and coordinate AI workloads externally.
The success of this strategy will depend on whether enterprises prefer database-centric automation or prefer keeping orchestration logic separate from data storage. Oracle is clearly betting that reducing the number of layers between agents and data will win on performance and security grounds—time will tell if that architectural opinion resonates with enterprise adoption patterns.