AgentDB Memory Patterns

This handles persistent memory for AI agents using AgentDB, which claims 150x-12,500x performance gains over traditional vector stores. You get session memory for conversations, long-term storage for facts and preferences, and pattern learning from successful interactions. The CLI setup is straightforward (npx agentdb init, then wire it into Claude with MCP), and it includes templates for different reinforcement learning approaches like Q-learning and actor-critic. The real value is in hierarchical memory organization and consolidation strategies that let you prune low-value memories over time. Use this when building chatbots or assistants that need to remember context across sessions without rebuilding state from scratch every time.

Install to Claude Code

npx -y skills add spencermarx/open-code-review --skill "AgentDB Memory Patterns" --agent claude-code