OceanDB field notes
AI agent memory, without the fog.
How persistent memory works, where RAG stops, what MCP changes, and why the memory your agents share should remain yours.
Foundations
- What is AI agent memory? Types, architecture, and examples A practical model for how agents retain useful context across turns, sessions, tools, and time. AI agent memory
- Persistent memory for AI agents: a practical guide What persistent memory must preserve, how the main architectures differ, and how to choose one. persistent memory for AI agents
- AI agent memory vs. RAG vs. context windows Three ways to give an agent context, the jobs each one does well, and why they often belong together. AI agent memory vs RAG
Implementation
- One memory for Claude, ChatGPT, Cursor, and Codex How a shared MCP memory keeps context portable across the AI tools you already use. memory across ChatGPT Claude Cursor Codex
- MCP memory server: how persistent agent memory works The protocol, authorization, tools, and data model behind memory that works across MCP clients. MCP memory server
- Persistent memory for AI coding agents The project context coding agents should retain across sessions—and what should remain in the repository. persistent memory for AI coding agents
Ownership
- Who owns your AI's memory? Memory contains the durable model of a person or team. Ownership must include access, correction, portability, and control over maintenance. AI memory privacy and ownership
- How AI memory handles contradictions and forgetting Why changing facts need provenance, history, and review rather than a blind overwrite or an endless archive. AI agent memory contradictions and forgetting
Evaluation
- The best AI agent memory system depends on the work A criteria-led guide to evaluating memory layers without pretending one architecture wins every workload. best AI agent memory systems
- Build or buy an AI agent memory layer? A decision framework for choosing between database primitives, an open-source stack, and a managed memory service. build vs buy AI agent memory
- SQL-backed AI memory vs. vector-only memory Why embeddings improve recall but do not replace structure, constraints, provenance, or exact queries. SQL AI memory vs vector database