ARMIN

The README opens with

Agent Reasoning Memory & Introspection Network. Queryable decision memory for AI agents.

ARMIN monitors AI agents as they work. It watches tool calls and prose, figures out what actually matters for future sessions, and builds a small graph of decisions and open items as the session unfolds. When the session is over — or when context compaction wipes the "why" — that reasoning survives, queryable and injectable.

AI agent sessions produce decisions. They also produce the context that explains those decisions — constraints nobody wrote down, questions that were raised but never answered, choices that bind future work. That context lives in the agent's context window, or in a log nobody reads, or nowhere at all.