Enterprise AI Agents are in General Availability in OCI Generative AI

OCI Generative AI now offers Enterprise AI agents that speed up the development and deployment of production-grade agentic applications through modular and composable primitives.

OCI Responses API

OpenAI Responses-compatible unified API for interacting with models and agentic tasks, including orchestration, reasoning, tool use, and memory. Supports multi-model routing and compatibility with existing OpenAI SDKs.

• Supported Models
• OpenAI: openai.gpt-oss-20b, openai.gpt-oss-120b
• xAI: xai.grok-3, xai.grok-3-mini, xai.grok-3-fast, xai.grok-3-mini-fast, xai.grok-4, xai.grok-4-fast-reasoning, xai.grok-4-fast-non-reasoning, xai.grok-4-1-fast-reasoning, xai.grok-4-1-fast-non-reasoning, xai.grok-code-fast-1
• Google: google.gemini-2.5-flash, google.gemini-2.5-flash-lite, google.gemini-2.5-pro

Tools

OpenAI-compatible integrations for File Search, Code Interpreter, Function Calling (local tools), and MCP Calling (remote MCP servers), Containers API, Vector Stores API, and Files API for building custom retrieval, storage, and execution workflows.

Memory

OpenAI-compatible features for conversation state management, including Conversations API, long-term memory, and short-term memory context compaction for all models supported through the OCI Responses API.

Projects New resource model for organizing agent workloads, isolation of conversations, files, containers and memory by project, configurable data retention and memory settings, replaces the conversation store used during the limited availability (LA) offering of the feature. Applications Fully managed hosting capabilities for agentic applications build with OSS Frameworks or MCP Servers with build in security and public and private endpoint support. Vector Stores

Managed vector storage with support for file ingestion, semantic search and metadata filtering enabling RAG and NL2SQL use cases.

NL2SQL

Fully solution to ingest customer schema information, enrich into a semantic vector store, and allow for natural language queries that produce SQL and run against the customer database in a permission controlled manner without moving or copying database content.