Oracle AI Database Private Agent Factory 26.4 is now available

Oracle AI Database Private Agent Factory 26.4 introduces new capabilities that help teams build, connect, observe, and deploy enterprise AI agents with more control.

This release expands support for agentic research, enterprise content ingestion, structured data enrichment, workflow orchestration, observability, and deployment portability. It is designed for teams that want to move beyond demos and build AI agents that work with trusted enterprise data, fit into existing processes, and can be operated in production environments.

The major additions in this release include:

  • Deep Data Research Agent for enterprise document ingestion and cited, context-aware answers
  • AI Enrichment using annotations to add business context to database objects
  • Amazon S3-compatible object storage as a new data source for knowledge-based agents
  • Knowledge Agent as a callable tool node in Agent Builder
  • Agent observability integrations with Arize Phoenix, Comet Opik, and Langfuse
  • Workflow export and import for easier migration across environments
  • New Agent Builder nodes for PL/SQL, Slack webhooks, and deterministic MCP tool execution
  • Security updates across Agent Factory components

Why this release matters

Enterprise AI agent projects often start with a simple goal: help employees answer questions, automate repetitive work, or make better use of trusted business data.

The harder part is making those agents useful in real operating environments.

Teams need agents that can search approved enterprise content, cite sources, connect to structured data, call business logic, notify downstream systems, and be monitored when something goes wrong. They also need a path to move agents from development and demos into production environments.

Oracle AI Database Private Agent Factory 26.4 focuses on those practical needs.

It gives customers more ways to connect agents to enterprise knowledge, enrich structured data with business context, orchestrate multi-step workflows, trace agent behavior, and manage deployment across environments.

What is Oracle AI Database Private Agent Factory?

Oracle AI Database Private Agent Factory helps teams build AI agents and agentic workflows using Oracle AI Database and enterprise data sources.

It provides a no-code and low-code environment for creating agents, connecting them to data, configuring workflows, and deploying AI-powered experiences that are grounded in trusted information.

With this release, Agent Factory adds more capabilities for agentic RAG, data enrichment, observability, and operational workflow design.

What’s new in Oracle AI Database Private Agent Factory 26.4?

1. Deep Data Research Agent

The new Deep Data Research Agent helps teams build research-style agents over approved enterprise documents.

It can ingest enterprise documents, prepare searchable knowledge bases, materialize retrieval tools, and return cited, context-aware answers through configurable agentic RAG workflows.

This is useful for teams building agents for: internal research, policy and procedure lookup, support enablement, field and sales enablement, and knowledge-intensive operations.

Instead of manually searching across multiple repositories, users can ask questions and receive grounded answers based on approved content.

2. AI Enrichment using annotations

AI Enrichment now supports annotations, making it easier to add and manage business context for structured data.

Users can add annotations to supported schema objects, including: schemas, tables, views, packages, functions, procedures, and materialized views.

This helps teams make structured enterprise data more understandable to AI workflows.

For example, a database object may have a technical name that is clear to developers but not obvious to an AI agent or business user. Annotations allow teams to capture trusted metadata centrally so it can be reused across agents, applications, microservices, and downstream AI experiences.

3. Amazon S3 as a data source

Oracle AI Database Private Agent Factory 26.4 adds Amazon S3 as a supported data source.

Teams can now create and manage S3 data sources by configuring bucket settings, credentials, filters, and crawl options. This allows agents to use S3-hosted enterprise content without requiring teams to manually move documents into the application.

The release also supports include filters, exclude filters, and excluded file extensions, giving teams more control over which S3 content is made available to agents.

4. Knowledge Agent as a tool node in Agent Builder

Agent Builder now supports Knowledge Agent as a tool node.

This makes it easier to bring trusted knowledge search into visual agent workflows. A configured Knowledge Agent can be used as a callable tool or as an independent step that produces a message for downstream workflow logic.

Users can configure: search query, language, result limit, and manual prompts or prompts passed from upstream workflow output.

This helps teams combine grounded knowledge retrieval with multi-step business workflows.

5. Agent observability and tracing

Oracle AI Database Private Agent Factory 26.4 adds agent observability support to help users and operators understand how agents behave.

Teams can trace activity across: LLM calls, tool invocations, agent turns, flow steps, and workflow execution paths.

The release supports OpenTelemetry-compliant tracing providers that can be configured directly in the UI, including Arize Phoenix, Comet Opik, and Langfuse.

This gives teams more visibility into prompts, agent runs, tool usage, and execution behavior, which can help with debugging, performance analysis, and operational tuning.

Masking options are also available to help teams troubleshoot safely.

6. Workflow export and import

Agent Builder now supports workflow export and import.

This helps teams move custom workflows across environments, back up configurations, and promote flows from development or demo environments into production.

Users can export selected or all owned custom workflows from My Custom Flows into a portable .paf file. Exported workflows are protected with a required password before they can be shared or imported into another environment.

This is especially useful for:

  • Environment migration
  • Demo setup
  • Tenant transfer
  • Backup and restore workflows
  • Production promotion

7. New Agent Builder nodes

Oracle AI Database Private Agent Factory 26.4 adds new built-in nodes that help teams connect agents to databases, collaboration tools, and MCP-based integrations with less custom code.

PL/SQL Tool Node

The PL/SQL Tool Node lets workflows execute selected Oracle PL/SQL procedures and functions directly from Agent Builder.

This allows teams to connect agent workflows to controlled database logic using discovered routines and bound arguments.

Slack Webhook Tool Node

The Slack Webhook Tool Node allows workflows to send results, alerts, summaries, and status updates directly to Slack channels through an incoming webhook.

This is useful for workflows that need to notify teams, route approvals, or share operational updates.

Deterministic MCP Tool Node

The Deterministic MCP Tool Node allows workflows to invoke a specific tool from a saved MCP server using explicit JSON input.

This supports predictable, non-agentic tool execution inside a workflow, which is useful when teams want controlled tool behavior as part of a larger agentic process.

8. Security updates

This release includes security updates across Agent Factory components.

These updates include dependency upgrades, operating system patches, vulnerability fixes, and security hardening measures designed to improve the overall security posture of Agent Factory deployments.

Customers should upgrade to 26.4 to benefit from the latest security improvements.

What customers can build with 26.4

Oracle AI Database Private Agent Factory 26.4 supports several practical enterprise AI agent patterns.

Research and knowledge agents

Teams can create agents that search approved enterprise documents and return grounded, reviewable answers. This is useful for policy lookup, support research, internal enablement, and document-heavy workflows.

Structured data agents

With AI Enrichment annotations and existing Select AI foundations, teams can add business context to database objects and use that context in agents, workflows, and downstream AI experiences.

Operational workflow agents

With new Agent Builder nodes, teams can connect agents to approvals, notifications, database logic, external tools, and multi-step workflow orchestration with less custom integration work.

Production-ready agent operations

With workflow export and import, application tracing, masking options, proxy configuration, and broader Linux support, teams can more easily move agents from demo environments into production operations.

Example enterprise use cases

Oracle AI Database Private Agent Factory is already being applied to practical enterprise scenarios, including contract intelligence, invoice anomaly detection, and diagnostic research assistance.

Examples include:

  • A contract intelligence and accountability agent connected to enterprise contract, invoice, and procurement systems
  • An invoice anomaly detection agent that reviews invoices before billing runs are finalized
  • An intelligent diagnostic research assistant that helps support teams investigate complex operational incidents

These patterns reflect a common theme: agents are most useful when they are grounded in trusted data and connected to the workflows where employees already work.

Get hands-on with Oracle AI Database Private Agent Factory

Teams that want to explore Agent Factory can start with the LiveLab hands-on experience.

The LiveLab walks users through how to install and configure Agent Factory, build a first agent, bootstrap agents from templates, and customize behavior using the Agent Builder UI.

No deep expertise is required. Users need access to OCI and can launch the demo environment to get started.

Availability

Oracle AI Database Private Agent Factory 26.4 is available through:

  • oracle.com for manual installation on Linux and macOS, across x86_64 and ARM64 architectures
  • OCI Marketplace for a faster launch experience using the marketplace listing
  • Product documentation for setup, configuration, and feature guidance

Learn more

To learn more about Oracle AI Database Private Agent Factory 26.4, visit the product documentation, download the release, or launch from OCI Marketplace.

For feature requests, roadmap feedback, or integration discussions, customers can contact the Oracle AI Database Private Agent Factory product management team.

FAQs

What is Oracle AI Database Private Agent Factory 26.4?

Oracle AI Database Private Agent Factory 26.4 is a release that adds new capabilities for building, integrating, observing, and deploying enterprise AI agents. It includes Deep Data Research Agent, Amazon S3 support, AI Enrichment annotations, workflow export and import, new Agent Builder nodes, observability integrations, and security updates.

What is the Deep Data Research Agent?

Deep Data Research Agent is a pre-built agent capability that ingests enterprise documents, prepares searchable knowledge bases, and returns cited, context-aware answers through configurable agentic RAG workflows.

Does Oracle AI Database Private Agent Factory support Amazon S3?

Yes. Oracle AI Database Private Agent Factory 26.4 adds Amazon S3 as a supported data source, allowing teams to ingest and retrieve from S3-hosted enterprise documents.

What observability tools are supported in Agent Factory 26.4?

Oracle AI Database Private Agent Factory 26.4 supports agent observability integrations with OpenTelemetry-compliant providers, including Arize Phoenix, Comet Opik, and Langfuse.

Can Agent Builder workflows be moved between environments?

Yes. Agent Builder now supports workflow export and import using portable .paf files protected by a required password. This helps teams migrate, back up, and promote workflows across environments.

What new Agent Builder nodes are included in 26.4?

The release adds new Agent Builder nodes for PL/SQL, Slack webhooks, Knowledge Agent, and deterministic MCP tool execution.