claude-memory-kit
Health Uyari
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Low visibility — Only 6 GitHub stars
Code Gecti
- Code scan — Scanned 3 files during light audit, no dangerous patterns found
Permissions Gecti
- Permissions — No dangerous permissions requested
This provides a ready-to-use project structure that gives AI coding agents a persistent memory. It allows the agent to remember project context, decisions, and progress across different sessions, surviving context compression.
Security Assessment
The overall risk is Low. It does not request dangerous permissions and the automated code scan found no malicious patterns, hardcoded secrets, or dangerous behaviors. Since it is written in Shell, users should be aware that it executes local filesystem commands. However, the setup simply creates local directories and files (like a memory folder and journal). There are no signs of unauthorized network requests or external data transmission. Your project data remains strictly on your local machine.
Quality Assessment
The project is highly active, with its last push occurring today. It uses the standard permissive MIT license. The documentation is excellent, clearly explaining the setup and how the multi-project tracking system works. The only notable drawback is its low community visibility, currently sitting at only 6 GitHub stars. This means it has not yet been widely vetted by a large audience, so you are relying primarily on the creator's code rather than broad community testing.
Verdict
Safe to use.
Ready-to-use project structure for Claude Code agents — memory, context, skills, hooks. Based on 1000+ sessions.
Give Claude Code a memory that grows with your projects.
Problem · Start · What You Get · How It Works · FAQ
The Problem
Every time you start a new chat with Claude, it forgets everything.
Your project structure. The decisions you made last week. That bug you fixed yesterday. The naming conventions your team follows. Gone.
You spend the first 10 minutes of every session catching the AI up on context it already had — and lost.
Claude Memory Kit solves this in a 5-minute setup.
Your agent remembers what it learned. It tracks your projects. It picks up exactly where it left off — even after the conversation compresses or you close the terminal.
What You Get
An agent that learns and remembers
Claude keeps a structured notebook (.claude/memory/) that persists between sessions. Patterns it discovers, decisions you make together, things that worked and things that didn't — all saved automatically and loaded next time.
The memory grows with your project. After 10 sessions, your agent knows your codebase better than a new team member.
Multi-project tracking
Each project gets its own Journal — a single file where Claude tracks tasks, decisions, and progress. Run multiple projects in parallel. Each one has isolated context, so work on Project A never leaks into Project B.
Context that survives compression
Long conversations get compressed by Claude Code. Without this kit, your progress disappears. With it, a built-in safety net saves everything before compression kicks in. Next session starts exactly where you left off.
Structured research
Not sure which approach to take? The experiments system gives Claude a framework for structured investigation: define a question, explore options, compare results, make a GO/NO-GO decision. Then create tasks from the result.
How to Start
git clone https://github.com/awrshift/claude-memory-kit.git my-project
cd my-project
claude
Claude handles the rest. It will ask your name, project name, and preferred language — then set everything up.
Requirements: Claude Code + a Claude subscription or API key.
How It Works
Five components, one system
| Component | File | What it does |
|---|---|---|
| Brain | CLAUDE.md |
Agent's identity, behavior rules, workflow instructions |
| Memory | .claude/memory/ |
Patterns and knowledge that persist across sessions |
| Rules | .claude/rules/ |
Domain-specific behavior (auto-loaded every session) |
| Journal | projects/X/JOURNAL.md |
Tasks, decisions, and status for each project |
| Context Hub | context/next-session-prompt.md |
"Pick up here" note between sessions |
What loads when
Not everything loads every time. The system uses layers — heavy context only loads when needed:
| Layer | What | When |
|---|---|---|
| Always on | Brain + Rules + Memory index | Every session |
| Session start | "What to do next" prompt | First thing |
| On demand | Project journal, topic files, experiments | When you work on them |
The session cycle
Every session: load context, do work, save progress. Automatic hooks make sure nothing is lost — even if you forget to save.
After 10 Sessions
Here's what changes as the memory accumulates:
| Session 1 | Session 10 |
|---|---|
| You explain your project from scratch | Agent already knows the codebase, conventions, and history |
| Generic suggestions | Suggestions based on what actually worked before |
| No awareness of past decisions | References previous decisions and their outcomes |
| Starts cold every time | Picks up mid-task with full context |
The kit doesn't just remember — it learns what matters. Verified patterns get promoted to permanent memory. One-off details stay in session notes and naturally fade.
Extending with Skills
The kit focuses on memory and context. For additional capabilities, install community skills into .claude/skills/:
| Skill | What it does | Repo |
|---|---|---|
| Gemini | Second opinions from Google AI | awrshift/skill-gemini |
| Brainstorm | Claude x Gemini structured debate | awrshift/skill-brainstorm |
| AWRSHIFT | Decision framework with evidence | awrshift/skill-awrshift |
FAQ
Do I need to know how to code?No. After setup, you talk to Claude in plain language. "Read the marketing plan and draft three emails" works just as well as technical commands.
Is my data private?Yes. Everything stays on your computer. Claude Code talks to Anthropic's API, which does not train on your data by default.
How much does it cost?The kit is free and open source. Claude Code needs a Claude Max/Pro subscription ($20-100/mo) or Anthropic API credits (pay-per-use).
Can I use this with an existing project?Yes. During setup, choose "I have existing code" and point Claude to your codebase. It will analyze the structure and set up context around it.
What happens if I mess up the memory files?Everything is plain text in git. Roll back with git checkout .claude/memory/ or just delete and let Claude rebuild from your project files.
Yes. Each project gets its own Journal and context section. Multiple Claude Code windows can work on different projects simultaneously without conflicts.
Built by Serhii Kravchenko · MIT License
Yorumlar (0)
Yorum birakmak icin giris yap.
Yorum birakSonuc bulunamadi