The Oracle AI Optimizer and Toolkit (the AI Optimizer) is an agentic AI platform for designing, running, evaluating, and refining GenAI applications. It gives developers and data scientists a practical workbench for model-driven agents and flows.
Agents can use configured tools to retrieve information or act on external systems, while flows compose model and tool steps into a controlled sequence. The built-in routes ground responses in private structured and unstructured data through Oracle AI Database Vector Search and SQLcl MCP, using Retrieval-Augmented Generation (RAG) and Natural Language to SQL (NL2SQL).
Grounding gives the runtime relevant application data at request time, which can improve accuracy and reduce reliance on a model's training knowledge.
Agentic Runtime
The Chatbot selects a runtime route from the tools enabled in Tool Selection:
| Tool selection | Runtime route | Behavior |
|---|---|---|
| No tools | LLM-only Agent | Generates a response with the selected language model and conversation context. |
| NL2SQL | NL2SQL Agent | Selects from a restricted SQLcl MCP tool inventory to query structured data. |
| Vector Search | Vector Search Flow | Runs the configured retrieval steps and uses the results to generate a grounded answer. |
| Both tools | Combined Router | Classifies the request, runs one route or both routes, and can synthesize their results. |
The reusable AgentSpec definitions and their LangGraph runtime are described in Agents and Flows.
*Project names and logos are shown for identification purposes only. Unless expressly stated, their use does not imply affiliation with, sponsorship by, or endorsement by Oracle or the respective project owners.
Architecture
The Oracle AI Optimizer and Toolkit (the AI Optimizer) consists of an API Server (AI Optimizer Server) and an optional web client (AI Optimizer Client) component. The server hosts the REST API, agent runtime, and MCP endpoint. Both the API Server and client can run directly on a developer workstation, in on-premises infrastructure, or in cloud environments.
Depending on the capabilities you plan to use, you may also need the following components, which are not delivered with the AI Optimizer. These can be local, on premises, or cloud-hosted:
- Oracle AI Database, including Oracle AI Database Free
- Access to at least one language model
- Access to at least one Embedding Model (for Retrieval Augmented Generation)
- (Optional) Access to an SQLcl MCP Server-enabled instance
The AI Optimizer supports Kubernetes deployments using a Helm chart. For deployment guidance, see Kubernetes / Helm.
Features
The AI Optimizer brings together the capabilities you need to build, test, and refine agentic GenAI solutions on the Oracle AI Database: