使用方法
- 复制安装命令,在自己的智能体环境安装前先检查源码包。
- 从原始文档核对所需工具、凭据和支持的宿主。
- 说明输入材料与期望交付物,并对照来源检查输出。
npx skills add https://github.com/smartworldarafath/rea_skill --skill claude-mem开始前需了解
- 安装不包含模型订阅、API 额度或访问文件的权限。
- 仓库星数和安装量反映活跃情况,不等于独立质量或安全测试。
按源仓库内容呈现,保留标题、案例、代码、表格、链接以及原文引用的演示图片。
Claude-Mem
Persistent memory compression system built for Claude Code.
What It Does
Claude-Mem compresses conversation transcripts and project context into compact memory snapshots that persist across sessions, enabling:
- Cross-session memory: Claude remembers what was done in previous sessions without manually re-loading context
- Knowledge graph compression: Project knowledge is distilled into structured memory entries
- Team cowork: Share memory across multiple agents working on the same project
- Multi-tool support: Works with Claude Code, Cursor, Grok Bot, Windsurf, OMP, Hermes, Pi, and more
Installation
npx claude-mem installOr via the Claude Code plugin marketplace:
/plugin marketplace add thedotmack/claude-memHow It Works
- Compress: At the end of a session, Claude-Mem analyzes the conversation transcript and extracts key decisions, code changes, and context
- Store: Compressed memory is stored as structured YAML/JSON in your project's
.claude-mem/directory - Load: On the next session start, memory is loaded and injected into Claude's context window
- Cowork: Multiple agents can read and write to the shared memory store
Commands
/mem compress # Compress current session into memory
/mem load # Load project memory at session start
/mem status # Show what's in memory
/mem search <query> # Search memory for specific information
/mem forget <topic> # Remove memory entries
/mem cowork # Enable multi-agent memory sharingConfiguration
Memory is stored in .claude-mem/ in your project root (add to .gitignore if private). Configure in .claude-mem/config.json:
{
"compression": "aggressive",
"max_entries": 100,
"include": ["decisions", "code_changes", "context"],
"cowork": false
}Reference
- Source: github.com/thedotmack/claude-mem
- License: Apache 2.0