Oracle AI Database Private Agent Factory
Deploy governed AI agents where your enterprise data already lives. Build, test, and run private agents that use Oracle Database, enterprise documents, APIs, and approved models while keeping security, identity, and deployment control with your organization.
What you can build with Private Agent Factory
Create governed agent workflows
Agent Builder provides teams with a no-code platform to integrate models, enterprise knowledge, Oracle Database, Select AI, APIs, and MCP tools, enabling the building, testing, and reusing of workflows with embedded governance
Turn enterprise knowledge into trusted answers
RAG Agent delivers enterprise-ready RAG using trusted content from SharePoint, websites, file systems, Amazon S3-compatible sources, and other approved repositories. Answers are grounded in source-linked knowledge users can verify.
Explore data and visualize results
Utilize Data Analysis Agent to explore Oracle Database tables, ask natural-language questions, and create visualizations; for business queries requiring shared definitions, rules, and SQL review, explore Deep NL2SQL
Synthesize complex research from enterprise files
Deep Research RAG Agent assists teams in exploring complex questions by retrieving relevant content, synthesizing findings, and providing cited answers from shared files and enterprise documents
Deep NL2SQL: From Business Questions to Trusted SQL
The agent clarifies before generating SQL
Ask complex business questions in natural language with Deep NL2SQL, which uses your Business Knowledge Graph to link your queries to the correct data, definitions, and rules. When a question has more than one possible meaning, the agent asks for clarification before generating SQL.
Give SQL the business context it needs
Connect your data with shared business definitions, metrics, and rules in a Business Knowledge Graph. Give agents reusable context to interpret questions consistently, select relevant data, and generate SQL that reflects how your business works.
Evaluate knowledge changes before publishing
Test how updates to your Business Knowledge Graph affect SQL generation. Compare generated SQL with reference queries, identify gaps, and review results before publishing changes, helping teams improve answer quality with confidence.
Make business context reusable
Capture business meaning, relationships, and rules alongside database objects and columns with Annotation Workbench. Reuse this context across agents to help them interpret data consistently and reduce the need to explain the same business logic repeatedly.
What's new in Private Agent Factory
Explore the latest capabilities for asking business questions, connecting agents to applications, and operating with enterprise control.
Deep NL2SQL with business knowledge
Ask business questions using company definitions, metrics, rules, and examples. A Business Knowledge Graph, clarification, SQL review, and evaluation help teams refine answers. Knowledge, Memory, and Annotation Workbenches maintain shared context.
Agent memory and multimodal workflows
Carry useful context across conversations with Agent Memory. Combine images, documents, web content, data analysis, and Jira in workflows that work with more than text.
Developer integration and agent collaboration
Bring agents into applications with the Python SDK and standardized REST APIs. Connect compatible agents through A2A, with JSON and Markdown content and asynchronous task handling.
Reusable knowledge preparation and templates
Prepare documents and content with reusable ingestion services, then use that knowledge in research and custom workflows. Start from templates to reuse proven agent patterns.
Native observability
Investigate agent behavior through sessions, traces, latency, token usage, and errors. Move from an overall view to individual executions to troubleshoot and improve workflows.
Enterprise control and model choice
Govern shared resources with users, groups, roles, permissions, and API keys. Choose supported models from Oracle Private AI Services, OCI, Azure OpenAI, Anthropic, and OpenRouter, including private inference with Private AI Services Container.
Flexible deployment and scale
Use ready-to-deploy images from Oracle Container Registry and simpler CLI administration. Deploy on Kubernetes with Oracle Linux, Ubuntu, or Red Hat Enterprise Linux, and use Oracle Database Operator for Kubernetes for scale and high availability.
Business use cases
Compare invoice details with contract terms, purchase orders, vendor records, and exception policies so that agents can flag mismatches for review before payment decisions are made
Ask which products lead by revenue, units, or margin, and let Deep NL2SQL clarify the metric, apply business definitions, and generate SQL for review, enabling analysts to investigate the results
Investigate issues across runbooks, tickets, and approved enterprise documents using a research agent that connects relevant evidence and produces cited findings for expert review
How Private Agent Factory works
Connect data and business context
Integrate Oracle Database, enterprise documents, approved repositories, and APIs, utilizing annotations, memories, and Business Knowledge Graphs to ensure business terminology, rules, and examples are reusable across agents
Build, test, and run agents
Select a prebuilt agent or design a workflow in Agent Builder, then connect approved models and tools, test behavior, and utilize tracing, evaluation, and observability to continually enhance the agent
Enterprise control and flexibility
Empower platform, database, and security teams to control agent data access, model usage, and operations
Control data and access
Keep agents close to Oracle Database and approved enterprise sources while using SSO, users, groups, and role controls to govern access to shared resources
Choose models and integrations
Connect Oracle Database, PL/SQL, REST APIs, and MCP tools by selecting supported model providers and maintaining approved configurations under the platform team's control
Evaluate and observe agents
Evaluate agent behavior, review grounded responses and citations, and investigate sessions, latency, token usage, and errors using masked tracing and observability
Deployment paths
Deploy in OCI, on-premises, or supported multicloud environments with ease; start quickly with OCI Marketplace or download the package for greater control