claude-memory-kit

Manage layered memory for Claude Code agents across sessions.

28|7|Updated Mar 26, 2026
One-click install
npx skills add https://github.com/awrshift/claude-memory-kit --skill claude-memory-kit
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: claude-memory-kit
Source: https://github.com/awrshift/claude-memory-kit/tree/main
Command: npx skills add https://github.com/awrshift/claude-memory-kit --skill claude-memory-kit

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) components.

What problem does it solve?

This Skill solves the issue of context loss between sessions in Claude Code, ensuring that your agent remembers client preferences, project decisions, and established patterns across long-running workflows.

Core Features & Use Cases

  • Layered Memory: Automatically manages hot cache (MEMORY.md), session handoffs, and long-term knowledge articles.
  • Audit-Driven Promotion: Uses an end-of-session ritual to audit repeated patterns and promote them to canonical rules or knowledge.
  • Use Case: A consultant managing multiple clients can clone this kit per client, allowing the agent to instantly resume work with the correct brand voice, project history, and backlog status loaded automatically.

Quick Start

Run the close-session skill to audit today's work and generate a handoff note for your next session.

Frequently Asked Questions about claude-memory-kit

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I retain agent context across multiple Claude Code sessions?

Persistent memory for Claude Code agents is maintained through an audit-driven session workflow that manages a hot cache and session handoffs. This mechanism ensures cross-session context retention without external databases.

What is the best way to isolate memory when managing multiple projects in Claude Code?

Multi-project memory isolation is achieved by cloning the memory kit per client or project. This approach allows the agent to automatically resume with the correct project history and backlog status loaded without external databases.

How do I promote repeated agent patterns into long-term knowledge rules?

Repeated agent patterns are promoted to canonical rules by running an end-of-session audit ritual. This process audits today's work, detects recurring patterns, and promotes them to long-term knowledge articles.

Does Claude Code persistent memory require an external database to function?

Claude Code persistent memory does not require an external database to function. It implements cross-session context retention and multi-project isolation entirely through local files and hook-based state protection.

How do I prevent context rot and stale references in agent memory caches?

Context rot and stale references are prevented by implementing hook-based state protection and size-capped memory caching. These mechanisms automatically detect stale references and prevent uncontrolled context degradation.

What are the limitations of file-based agent memory compared to database-driven context management?

File-based agent memory relies on size-capped memory caching to prevent uncontrolled growth, imposing a strict upper bound on active context. Users must actively run session audits to manage stale references and maintain cache relevance.