About Agent Memory
Oracle AI Agent Memory provides a persistent memory layer for enterprise AI agents on the Oracle AI Database.
As AI systems evolve from single-turn assistants to autonomous, long-running agents, memory becomes foundational. Agent Memory enables agents to retain context across interactions, store useful information over time, and use that memory to make subsequent responses and actions more consistent and effective.
The SDK is designed around two core pillars: short-term memory and long-term memory.
Short-term Memory: Uses thread context cards and conversation summaries to maintain recent turns, task state, and intermediate progress during an active session. This is useful when an agent must preserve working context without sending the complete conversation to the model on every turn.
Long-term Memory: Uses add and search memory workflows to store and retrieve user preferences, learned rules, and facts from earlier interactions across sessions. These records can be searched with semantic, keyword, or hybrid strategies and scoped to the appropriate user, agent, and thread.
Agent Memory in the Agentic Ecosystem
Agent Memory is positioned as the governed memory core in the AI agents ecosystem. It gives agents durable context while leaving orchestration, authentication, authorization, and end-user identity management to the integrating application.
Agent Memory works with multiple LLM and embedding providers, while persisting enterprise memory on Oracle AI Database’s converged foundation. It can store and retrieve messages, memories, profiles, metadata, images, and linked memory records through one SDK.
For more information on compatible LLMs and Embedding Models, see Get Started with Agent Memory.
Key Benefits
Most memory stacks combine several specialized systems, increasing integration overhead and operational complexity. Agent Memory unifies these capabilities in a single enterprise platform, enabling teams to build stateful agents with improved consistency and security.
Unified Memory Core: Use Oracle AI Database as the converged platform for working memory and long-term memory, rather than piecing together separate vector, graph, JSON, and transactional stores.
Built for Enterprise: Leverage Oracle AI Database reliability and scalability to support persistent memory for real workloads, not just prototypes.
Platform Integration
Agent Memory can be used by Oracle AI Database Private Agent Factory by using an MCP Server. It can also be used with external frameworks such as LangGraph and WayFlow that need persistent, enterprise-grade memory.
For more information, see Use Agent Memory with an MCP Server.
Quick Start
To install the current Oracle AI Agent Memory release and configure its database prerequisites, see Get Started with Agent Memory. If you need a local Oracle AI Database for the examples, see Run Oracle AI Database Locally.
Next Steps
- Explore the API documentation to learn about the available classes, methods, and functions.
- Browse the how-to guides in the table of contents for task-focused walkthroughs and framework integrations.
- Review Security Considerations before enabling active-memory features or sharing a database connection across end users.
Security Considerations
Memory extraction and summarization require careful security assessment before deployment. The integrating application is responsible for end-user authentication, authorization, and correct memory scoping when calling the SDK.
For more information, see Security Considerations.