ReasoningBank With AgentDB

This implements the ReasoningBank pattern for agents that learn from experience, using AgentDB's vector database as the backend. You get trajectory tracking to record sequences of actions, verdict judgment to evaluate outcomes, and memory distillation to consolidate similar experiences into reusable patterns. The performance claims are aggressive (150x faster retrieval, sub-millisecond memory access) and it maintains backward compatibility with legacy ReasoningBank databases. Useful if you're building agents that need to remember what worked before, like optimization tasks or repeated problem-solving where past solutions inform future decisions. The API is straightforward: store patterns with embeddings, retrieve similar experiences, let the system consolidate memories over time.

Install to Claude Code

npx -y skills add spencermarx/open-code-review --skill "ReasoningBank with AgentDB" --agent claude-code