kevin-hs-sohn/hipocampus

Hipocampus Compaction builds a five-level memory compaction tree from raw logs into daily, weekly, monthly, and root indices. Use it at session start or from a scheduler to keep search indices current without deleting originals.

Hipocampus Core defines a multi-tier memory protocol for Claude Code sessions, including startup reads, compaction checks, checkpoints, and daily log updates. Use it when an agent needs persistent working state and structured memory maintenance.

Hipocampus Core defines a multi-tier agent memory protocol with session-start checks, task lifecycle rules, and compaction guidance. Use it to manage persistent memory files and structured checkpoints across sessions.

Hipocampus Core defines a 3-tier agent memory protocol with session start, checkpoint, and compaction rules. Use it to manage persistent memory files and task state across Claude Code sessions.

Hipocampus Core defines a 3-tier memory system for Claude Code, including session start checks, end-of-task checkpoints, and memory file maintenance. Use it to keep long-running agent work organized and recoverable.

Hipocampus Flush dumps the current session context to the daily raw log through a subagent. Use it before compaction when you want to persist what happened in the session.

Hipocampus Recall guides a three-step memory lookup: ROOT.md triage, manifest-based file selection, then qmd search fallback. Use it when a question may depend on past notes or compacted memory.

Hipocampus Search looks up memory with qmd using BM25 and optional vector search, then follows the compaction tree when needed. Use ROOT.md first to decide whether to search memory or look externally.