Oracle targets agentic use cases with AI Database 26ai

Oracle is doubling down its focus on agentic use cases for enterprises with new features added to its latest long-term support version of its database offering — AI Database 26ai. The new database includes Select AI Agent, Agent Factory, and integration with Oracle’s MCP servers.

AI Database 26ai, which has been made generally available, replaces the 2023-released Database 23c (later renamed to Database 23ai in 2024 after the addition of AI-based application development features, such as vector search).

The new database, according to Oracle, moves beyond supporting only generative AI use cases to also supporting agentic AI use cases and provides tools that will allow enterprises to build agents and agentic workflows, thereby driving automation.

“It’s about making AI Agents first-class citizens inside the database,” said Tony Baer, chief analyst at dbInsight.

Baer was specifically hinting at the Select AI Agent added to the database, which builds on Select AI — a feature inside Oracle Database introduced in 2023 that allowed users to query data in natural language.

Despite its nomenclature, Select AI Agent is a framework inside the database that will allow developers to build, deploy, and manage agents, Oracle said in a statement, adding that it will come with pre-built, in-database tools and external tools via REST and MCP servers.

“Select AI Agent is Oracle’s take on the emerging agent builder platforms…This is more of a pro-code experience where developers can more specifically target database functions and knowledge context integration,” said Alexander Wurm, principal analyst at Nucleus Research.

Oracle has also added another framework — AI Private Agent Factory — with similar capabilities to the database for non-technical users.

However, the functionality of this framework is limited to platform and security teams needing a no-code way to deploy private, containerized agents in controlled environments, said Stephanie Walter, practice leader of AI stack at HyperFRAME Research.

“Comparable offerings are Snowflake managed MCP plus partner kits or Google ADK patterns when self-hosting agents,” Walter added.

To further support agentic use cases, Oracle has also added MCP server support for the database that will allow developers to build agents or agentic workflows with the capability to interact with the data inside the database to answer queries.

Other features of the new database, which it mostly imbibes from its previous versions, include data annotations, vector search, Exadata support, and integration with Nvidia’s NeMo Retriever microservices.

In terms of security, the database comes with SQL Firewall, Zero Data Loss Cloud Protect, and quantum algorithms to encrypt data in motion and at rest.

While the firewall protects against unauthorized SQL activity and injection attacks, the Zero Data Loss Cloud protects on-premises Oracle databases from data loss and ransomware using the Zero Data Loss Recovery Service.

Existing customers of Database 23ai can move to the newly released 26ai version simply by applying the October 2025 update, without the need for a full database upgrade or application re-certification, Oracle said. AI Database 26ai also forms the base for Oracle’s new Autonomous AI Lakehouse that supports Apache Iceberg for interoperability, it added.