Acteq1391gp/claude-code-memory-guide

Guide to building file-based persistent memory for Claude Code sessions.

What it does

• CLAUDE.md config — central file with identity, rules, style, and behavioral constraints Claude reads every session
• MEMORY.md index — table of contents for all memory files; automatically loaded to know what context is available
• Memory file architecture — organized by type (user, feedback, project, reference) with YAML frontmatter and markdown body
• Changelog/journal — running log of decisions and actions with timestamps and details
• Wiki compilation — pattern for consolidating scattered memories into topic-based reference documents

How it works

Four-layer stack: CLAUDE.md (core config) → MEMORY.md (index) → memory/ directory (individual files by type) → changelog (action journal). Memory files use YAML frontmatter to declare name, type, and description; Claude reads MEMORY.md on every session start to load relevant context files. The system compiles when accumulated memories on a topic reach 50+ files, replacing scattered entries with organized wikis. Built on the principle that Claude can retain context across sessions if given persistent, organized structure to read.

Standing & signals

Single-author guide published May 2026 (3.5 months old). 1 star and no recent activity indicate minimal adoption and community pickup. The README claims production use of 220+ memory files across 18 categories, suggesting the author uses this system extensively but has not attracted external adoption. Guide quality is high (detailed examples, templates, ASCII diagrams), but it remains a personal productivity system without backing org or active community.

Q2 — Apply where

Not-now. User is already running this exact pattern (evidenced by existing memory structure at ~/.claude/projects/*/memory/ with MEMORY.md index and typed memory files). Adopting this guide adds no value since the workflow is already operational and integrated into current sessions.

Q3 — Fit for current projects + sketch

No direct fit for active projects. Research Console and AI Learning Path do not need memory system setup (already in place). ClientCo Intelligence Line and Financial Analysis operate as independent tools without Claude Code session memory. The guide would only be valuable if the user needed to teach this system to others or validate/upgrade their existing memory patterns—neither is an active requirement.

Q4 — Future-watch verdict

IGNORE. The guide is well-written but addresses a solved problem (user's memory system is operational). No adoption signals (1 star) and 3.5 months without updates indicate static documentation. No reason to track further unless explicitly building Claude Code onboarding materials for others.