使用说明

xMemory 四层记忆架构 · 智能分类 · 自动提取 · 多 IDE 支持

🎯 Introduction

Tired of having to repeat self introductions every time you switch AI tools?

Cursor doesn't know that you mentioned in Claude that you like TypeScript, Augment doesn't remember your technical decision yesterday ..

OpenMemory Plus allows all AI agents to share the same memory.

OpenMemory Plus It is a unified memory management framework designed for AI agents, using XMemory four layer memory architecture：

L3 Theme - Theme clustering layer, automatically summarizing high-level concepts
L2 Semantic - Semantic memory layer, based on BGE-M3 vector retrieval
L1 Episode - Plot memory layer, storing dialogue context
L0 Message - Original message layer

# 5 分钟安装，终结 AI 失忆症
npx openmemory-plus install

📸 Effect Display

_{自动记忆提取执行}

_{记忆提取结果}

💡 Why do we need OpenMemory Plus?

Pain point scenario

Have you encountered these problems before?

😤 Pain points	📖 Scene description
Multiple CLI memory fragmentation	Simultaneously using Gemini, Augment, Claude, and cursor, each with independent memories that are not interconnected
Agent amnesia	Every time there is a new conversation, the agent forgets who you are, where the project is, and what you did last time
Repeat self introduction	After switching CLI, I have to tell the Agent: I like TypeScript, use pnpm, and prefer functional programming ..
Fragmentation of configuration	Deploy URL in Slack, API Key in notes, path in mind
Context loss	I don't remember the technical decision discussed by Claude yesterday, but I don't remember it at all in Augment today

How to solve OpenMemory Plus?

🔗 Multi CLI Unified Memory Layer

┌─────────────────────────────────────────────────────────────┐
│                    没有 OpenMemory Plus                     │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  ┌─────────┐   ┌─────────┐   ┌─────────┐   ┌─────────┐     │
│  │ Gemini  │   │ Augment │   │ Claude  │   │ Cursor  │     │
│  │ CLI     │   │ Agent   │   │ Code    │   │ Agent   │     │
│  ├─────────┤   ├─────────┤   ├─────────┤   ├─────────┤     │
│  │ 记忆 A  │   │ 记忆 B  │   │ 记忆 C  │   │ 记忆 D  │     │
│  │ (独立)  │   │ (独立)  │   │ (独立)  │   │ (独立)  │     │
│  └─────────┘   └─────────┘   └─────────┘   └─────────┘     │
│       ❌ 互不相通，每个都要重复告知偏好和上下文              │
│                                                             │
└─────────────────────────────────────────────────────────────┘
                            ↓
┌─────────────────────────────────────────────────────────────┐
│                 使用 OpenMemory Plus                        │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  ┌─────────┐   ┌─────────┐   ┌─────────┐   ┌─────────┐     │
│  │ Gemini  │   │ Augment │   │ Claude  │   │ Cursor  │     │
│  │ CLI     │   │ Agent   │   │ Code    │   │ Agent   │     │
│  └────┬────┘   └────┬────┘   └────┬────┘   └────┬────┘     │
│       │             │             │             │           │
│       └─────────────┴──────┬──────┴─────────────┘           │
│                            │                                │
│                    ┌───────▼───────┐                        │
│                    │  OpenMemory   │                        │
│                    │  Plus 统一层  │                        │
│                    ├───────────────┤                        │
│                    │ • 用户偏好    │                        │
│                    │ • 技术栈      │                        │
│                    │ • 项目配置    │                        │
│                    │ • 历史决策    │                        │
│                    └───────────────┘                        │
│       ✅ 一处记忆，处处可用。切换 CLI 无缝衔接              │
│                                                             │
└─────────────────────────────────────────────────────────────┘

💬 Comparison of actual conversations

┌─────────────────────────────────────────────────────────────┐
│  [Gemini CLI] 早上                                          │
│  用户: 我喜欢用 TypeScript 和 pnpm                          │
│  Gemini: 好的，已记住你的偏好 ✅                            │
├─────────────────────────────────────────────────────────────┤
│  [Augment] 下午 - 切换 CLI                                  │
│  用户: 帮我初始化一个新项目                                 │
│  Augment: 好的！根据你的偏好，我用 TypeScript + pnpm 初始化 │
│           (自动读取了 Gemini 存储的偏好) ✅                 │
├─────────────────────────────────────────────────────────────┤
│  [Claude Code] 晚上 - 再次切换                              │
│  用户: 检查一下代码风格                                     │
│  Claude: 我看到你偏好函数式风格，检查中...                  │
│          (所有 CLI 共享同一份记忆) ✅                       │
└─────────────────────────────────────────────────────────────┘

🎯 Applicable scenarios

👤 Who is suitable for?

User Type	Usage Scenario
independent developer	Multi project development, hoping that the agent remembers the configuration and context of each project
Full Stack Engineer	Frequent switching between front-end and back-end projects requires the agent to remember the technology stack preferences
Heavy user of AI tools	Simultaneously using cursor, Claude, and Augment, hoping for mutual memory exchange
Team Tech Lead	I hope the project configuration will be version based, and new member agents will automatically obtain context

📋 Typical use cases

🔧 用例 1: 项目配置管理

sceneYou have 5 projects, each with different deployment configurations

Without OpenMemory Plus:

Always inform the Agent of the project path
Provide deployment URL every time
Explain the project structure every time

Using OpenMemory Plus:

<!-- _omp/memory/techContext.md (自动生成) -->
## 部署配置
- Vercel URL: https://my-app.vercel.app
- Project ID: prj_xxx

## 项目路径
- Root: /Users/me/projects/my-app
- Src: ./src
- Tests: ./tests

Agent automatically reads, no need for repeated explanation.

🎨 用例 2: 跨项目偏好同步

sceneDo you like to use TypeScript, Tailwind, pnpm

Without OpenMemory Plus:

Every new project must inform the agent of your preferences
Preference loss after switching IDE

Using OpenMemory Plus:

openmemory (用户级，跨项目共享):
├── "用户偏好: 使用 TypeScript 而非 JavaScript"
├── "用户偏好: 使用 pnpm 而非 npm"
├── "用户偏好: 使用 Tailwind CSS"
└── "用户技能: 熟悉 React, Next.js, Node.js"

Any project, any IDE, Agent knows your preferences.

📝 用例 3: 技术决策追踪

sceneAfter team discussion, it was decided to use PostgreSQL instead of MongoDB

Without OpenMemory Plus:

The decision record is in Slack/Motion, and the agent is unaware of it
Next time, the agent may suggest using MongoDB

Using OpenMemory Plus:

<!-- _omp/memory/techContext.md (自动记录) -->
## 技术决策

### 数据库选型 (2026-02-01)
- **选择**: PostgreSQL
- **备选**: MongoDB, MySQL
- **原因**: 需要复杂查询和事务支持

The agent remembers the decision and will not repeat the rejected proposal.

🔄 用例 4: 多 CLI 记忆共享 (核心场景)

sceneYou use Gemini CLI, Augment, Claude Code, and cursor simultaneously

Without OpenMemory Plus:

😤 每天的痛苦循环：

[早上 - Gemini CLI]
用户: 我喜欢 TypeScript，用 pnpm
Gemini: 好的！

[中午 - 切换到 Augment]
用户: 帮我创建组件
Augment: 请问你用 JavaScript 还是 TypeScript？  ← 又要说一遍
用户: TypeScript...
Augment: 用 npm 还是 yarn？  ← 又又要说一遍
用户: pnpm...  😤

[下午 - 切换到 Claude Code]
Claude: 你好！请问你的技术栈偏好是？  ← 又又又要说一遍
用户: ...... 😭

Using OpenMemory Plus:

✅ 一次告知，处处生效：

[早上 - Gemini CLI]
用户: 我喜欢 TypeScript，用 pnpm
Gemini: 好的，已记住！ → 存入 openmemory

[中午 - 切换到 Augment]
用户: 帮我创建组件
Augment: 好的！用 TypeScript + pnpm 创建中... ← 自动读取
         (我知道你的偏好 😊)

[下午 - 切换到 Claude Code]
用户: 检查代码风格
Claude: 根据你的 TypeScript 偏好检查中... ← 自动读取
        发现 3 处可优化 ✅

[晚上 - 切换到 Cursor]
Cursor: 我看到你今天在其他 CLI 创建了新组件，
        需要我帮你写测试吗？ ← 甚至知道你今天做了什么

Shared Memory Content:

# openmemory (所有 CLI 共享)
用户偏好:
  - 语言: TypeScript
  - 包管理器: pnpm
  - 框架: React, Next.js
  - 风格: 函数式编程

用户技能:
  - 熟悉: Node.js, Python
  - 学习中: Rust

今日上下文:
  - 在 Gemini 设置了偏好
  - 在 Augment 创建了 Button 组件
  - 在 Claude 优化了代码风格

🚀 Quick Start

One-click installation

npx openmemory-plus install

The installation wizard will automatically guide you:

✅ Detecting system dependencies (Docker, Qdrant, Ollama, BGE-M3)
✅ One click deployment of Docker Compose (Recommended) or native installation
✅ Select IDE type
✅ Initialize project configuration
✅ Display the next steps guide

🐳 Docker Compose One Click Deployment (Recommended)

Just install Docker, other dependencies will be automatically handled!

# 方式 1: 安装时自动检测并使用 Docker Compose
npx openmemory-plus install

# 方式 2: 显式使用 Docker Compose 模式
npx openmemory-plus install --compose

# 方式 3: 手动管理依赖服务
omp deps init      # 初始化配置
omp deps up        # 启动服务 (Qdrant + Ollama + BGE-M3)
omp deps status    # 查看状态
omp deps down      # 停止服务

Advantage:

🎯 Just install Docker as a dependency
⚡ One click start of all services
📦 BGE-M3 model automatic download
💾 Data persistence, restart without loss

Basic Usage

After installation, use:

/memory              # 显示记忆状态 + 快速菜单
/mem search <query>  # 搜索记忆
/mem sync            # 同步并检测冲突
/mem clean           # 清理过时记忆

System Requirements

Dependency	Version	Description
Node.js	>=18.0.0	Running CLI
Docker	最新版	The only necessary dependency (Docker Compose mode)
Ollama	Latest Version	Running BGE-M3 Embedded Model (Native Mode Required)

✨ feature

core competency

🧠 XMemory four layer architecture — L3 Theme → L2 Semantic → L1 Episode → L0 Message
🎯 Intelligent Classification - Automatically determine whether information should be stored in the project or user memory
🔍 semantic search - Multi language vector retrieval+topic level clustering based on BGE-M3
⚡ Event driven extraction - Automatically trigger memory retrieval skill at the end of the conversation
🔐 Sensitive information filtering - Automatically identify and prevent the storage of API keys, passwords, etc
🎯 Top-down search Starting from the theme layer, adaptively expanding to the semantic layer

🤖 Multiple LLM support ^NEW

Support multiple LLM providers for memory classification:

Provider	Model	Features
DeepSeek	deepseek-chat	🔥 Recommended, high cost-effectiveness
MiniMax	Abab6.5s-chat	Chinese optimization
ZhiPu	Glm-4-flash	Domestic Large Model
Qwen	Qwen turbo	Alibaba Cloud
OpenAI	Gpt-4o-mini	International Standard
Ollama	Local Model	Offline Available

🧠 Intelligent Memory Management

📊 Multi dimensional classification - Three dimensional classification system (Scope/Confidence/Temporal), accurately routing each memory
🧹 ROT intelligent filtering - Automatically identify redundant, outdated, and trivial information to keep the memory bank streamlined
⏰ Ebbinghaus attenuation - Memory lifecycle management based on forgetting curve, automatically clearing old memories
🔀 conflict detection - Automatically detect dual layer data conflicts and provide intelligent prompts for solutions
📈 Health monitoring Four dimensional weighted scoring, real-time monitoring of memory system status

📖 Technical DetailsSee: Memory System Architecture Document

Advanced features

📊 downgrade strategy - Automatically downgrade to local storage when MCP is unavailable
🔗 memory consolidation - Semantic clustering merging fragmented memory
📉 quality indicator - Visual panel displaying memory health status
🔧 Progressive configuration - Append rather than overwrite existing configuration files

🏗️ architecture

XMemory four layer memory architecture

┌─────────────────────────────────────────────────────────────┐
│                 xMemory 4-Layer Architecture                │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  ┌─────────────────────────────────────────────────────┐   │
│  │  L3 Theme (主题层)                                   │   │
│  │  ├── 自动聚类相似语义记忆                            │   │
│  │  ├── 主题吸附 (attachThreshold: 0.62)               │   │
│  │  ├── 主题分裂 (maxThemeSize: 12)                    │   │
│  │  └── 主题合并 (mergeThreshold: 0.78)                │   │
│  └─────────────────────────────────────────────────────┘   │
│                            ↑                                │
│  ┌─────────────────────────────────────────────────────┐   │
│  │  L2 Semantic (语义层) - Qdrant 向量数据库           │   │
│  │  └── BGE-M3 多语言 Embedding                        │   │
│  └─────────────────────────────────────────────────────┘   │
│                            ↑                                │
│  ┌─────────────────────────────────────────────────────┐   │
│  │  L1 Episode (情节层) - 对话上下文                   │   │
│  └─────────────────────────────────────────────────────┘   │
│                            ↑                                │
│  ┌─────────────────────────────────────────────────────┐   │
│  │  L0 Message (消息层) - 原始对话                     │   │
│  └─────────────────────────────────────────────────────┘   │
│                                                             │
└─────────────────────────────────────────────────────────────┘

Storage Tiering

┌─────────────────────────────────────────────────────────────┐
│                    OpenMemory Plus                          │
├─────────────────────────────────────────────────────────────┤
│                                                             │
│  ┌─────────────┐    ┌─────────────┐    ┌─────────────┐     │
│  │   /memory   │    │  /mem xxx   │    │  自动提取   │     │
│  │   (入口)    │    │  (子命令)   │    │  (Skill)    │     │
│  └──────┬──────┘    └──────┬──────┘    └──────┬──────┘     │
│         └──────────────────┼──────────────────┘             │
│                            ↓                                │
│  ┌─────────────────────────────────────────────────────┐   │
│  │              Memory Router (智能分类)                │   │
│  └─────────────────────────────────────────────────────┘   │
│                            │                                │
│         ┌──────────────────┼──────────────────┐            │
│         ↓                                     ↓            │
│  ┌─────────────────┐              ┌─────────────────┐      │
│  │ _omp/memory/   │              │   openmemory    │      │
│  │   (项目级)      │              │   (用户级)      │      │
│  ├─────────────────┤              ├─────────────────┤      │
│  │ • project.yaml  │              │ • L3 主题层     │      │
│  │ • decisions.yaml│              │ • L2 语义层     │      │
│  │ • themes/       │              │ • L1 情节层     │      │
│  │ • Git 版本控制  │              │ • MCP 协议      │      │
│  └─────────────────┘              └─────────────────┘      │
│                                                             │
└─────────────────────────────────────────────────────────────┘

classification rule

Information type	Storage location	Hierarchy	Example
Project Configuration	`_omp/memory/*.md`	-	Deploy URL, environment variables, path
Technical decision-making	`_omp/memory/techContext.md`	Framework selection and architecture design
Topic clustering	`_omp/memory/themes/`	L3	Automatically generated topic index
User preferences	`openmemory` (MCP)	L2	Language preference, code style
User Skills	`openmemory` (MCP)	L2	Familiar technology stack and experience

XMemory retrieval process

查询输入 → L3 主题层匹配 → 分数 > 0.75? → 展开到 L2 语义层 → 返回结果
                ↓                   ↓
           KNN 邻居搜索        直接返回主题

💡 attentionAfter installation, project level memory is stored in _omp/memory/ The directory will be added to Git version control.

🖥️ Supported IDEs

Augment
✅ 完整支持 Claude Code
✅ 完整支持 Cursor
✅ 完整支持 Gemini CLI
✅ 完整支持

📦 CLI commands

Installation command

# 交互式安装 (推荐)
npx openmemory-plus install

# 使用 Docker Compose 一键部署 (推荐)
npx openmemory-plus install --compose

# 静默安装
npx openmemory-plus install -y

# 指定 IDE
npx openmemory-plus install --ide augment

# 指定 LLM Provider (用于记忆分类)
npx openmemory-plus install --llm deepseek

# 仅配置项目，跳过依赖检测
npx openmemory-plus install --skip-deps

# 显示 MCP 配置
npx openmemory-plus install --show-mcp

Supported LLM Providers

Provider	Command	Environment Variables
DeepSeek	`--llm deepseek`	`DEEPSEEK_API_KEY`
MiniMax	`--llm minimax`	`MINIMAX_API_KEY`
by Zhi	`--llm zhipu`	`ZHIPU_API_KEY`
Qwen	`--llm qwen`	`DASHSCOPE_API_KEY`
OpenAI	`--llm openai`	`OPENAI_API_KEY`
Don't be	`--llm ollama`	(Local, no API Key required)

🐳 Dependency Service Management (Docker Compose)

# 初始化 Docker Compose 配置
omp deps init

# 启动所有依赖服务 (Qdrant + Ollama + BGE-M3)
omp deps up

# 启动前拉取最新镜像
omp deps up --pull

# 停止所有服务
omp deps down

# 查看服务状态
omp deps status

# 查看服务日志
omp deps logs              # 所有服务
omp deps logs ollama       # 指定服务
omp deps logs -f           # 持续输出

# 手动下载 BGE-M3 模型
omp deps pull-model

Diagnostic command

# 检查系统状态
npx openmemory-plus status

# 诊断问题
npx openmemory-plus doctor

# 自动修复
npx openmemory-plus doctor --fix

Agent built-in commands

Command	Description
`/memory`	Display quick status+sub command menu
`/mem status`	Detailed memory state
`/mem search {query}`	Semantic search memory (default L2 semantic layer)
`/mem search {query} --level theme`	🆕 Topic level search (L3)
`/mem search {query} --level theme --no-expand`	🆕 Only the theme layer, not expanded
`/mem store`	Manually storing memory
`/mem sync`	Detect and resolve conflicts
`/mem clean`	Clear ROT memory
`/mem decay`	Time decay analysis
`/mem graph`	Visualization of Knowledge Graph
`/mem themes`	🆕 View the status of topic clustering

📁 project structure

openmemory-plus/
├── cli/                   # CLI 工具源码
│   ├── src/
│   │   ├── commands/      # install, status, doctor
│   │   └── lib/           # 核心库
│   └── templates/         # 模板文件
│       └── shared/        # 共享模板
│           └── _omp/      # 核心目录模板
├── docs/                  # 文档
│   └── architecture.md    # 架构设计
├── AGENTS.md              # AI Agent 配置入口
└── README.md              # 本文件

# 安装后在你的项目中生成 (以 Augment 为例):
your-project/
├── _omp/                      # OpenMemory Plus 核心目录 (所有 IDE 共享)
│   ├── AGENTS.md              # 完整 Agent 规则文件
│   ├── memory/                # 项目级记忆存储
│   │   ├── projectbrief.md    # 项目概述
│   │   ├── productContext.md  # 产品需求
│   │   ├── techContext.md     # 技术栈
│   │   ├── activeContext.md   # 当前会话上下文
│   │   ├── systemPatterns.md  # 模式与规范
│   │   ├── decisions.yaml     # 架构决策日志
│   │   └── progress.md        # 任务进度
│   ├── commands/              # Agent 命令
│   │   └── memory.md          # 主命令入口
│   ├── workflows/             # 工作流
│   │   └── memory/            # 记忆管理工作流 (7 步骤)
│   └── skills/                # Agent Skills
│       └── memory-extraction/ # 记忆提取 Skill (自动触发)
│
├── AGENTS.md                  # 入口文件 (引用 _omp/AGENTS.md)
└── .augment/                  # IDE 特定目录
    ├── commands/              # 命令入口 (链接到 _omp)
    └── skills/                # Skills (链接到 _omp)

🔧 tech stack

Components	Technology	Description
CLI	TypeScript+Commander	Command Line Tools
Vector database	Qdrant	Store user memory vectors
Embedding	BGE-M3 (via Ollama)	Multilingual Text Embedding
Protocol	MCP (Model Context Protocol)	Agent Communication Protocol
Project Storage	YAML+Git	Version Control Configuration Files

⚙️ working principle

Memory retrieval process

对话/操作 → 信息检测 → 智能分类 → 结构化提取 → 双层存储
     │           │           │            │            │
     ▼           ▼           ▼            ▼            ▼
  用户输入    关键词匹配   项目 or 用户?   YAML 格式    _omp/memory/
  Agent 输出  模式识别    自动判断        去重去噪     openmemory

Classification Decision Tree

检测到有价值信息
        │
        ▼
   ┌─────────────┐
   │ 包含项目特定 │──是──▶ _omp/memory/ (项目级)
   │ 信息？      │        ├── techContext.md
   └─────────────┘        ├── productContext.md
        │ 否              └── activeContext.md
        ▼
   ┌─────────────┐
   │ 包含用户偏好 │──是──▶ openmemory (用户级)
   │ 或技能？    │        ├── 偏好设置
   └─────────────┘        ├── 技能经验
        │ 否              └── 对话上下文
        ▼
   ┌─────────────┐
   │ ROT 信息？  │──是──▶ 丢弃 (不存储)
   │ 冗余/过时/  │        ├── "好的"
   │ 琐碎       │        ├── "明白了"
   └─────────────┘        └── 临时状态

Sensitive information filtering

Automatically detect and阻止存储The following content:

Type	Detection Mode	Example
API Key	`sk-`, `api_key`, `token=`	`sk-xxx...`
Password	`password`, `secret`	`password: 123456`
Private key	`-----BEGIN`, `PRIVATE KEY`	SSH/GPG key
Personal information	ID card, bank card, mobile phone number	`13800138000`

🔄 Comparison with other schemes

Features	OpenMemory Plus	Pure OpenMemory MCP	Native Mem0 SDK	Manual. env	Notion/Document
Automatic extraction	✅ Automatic triggering at the end of the conversation	⚠️ Agent needs to actively call	⚠️ Code call add() is required	❌	❌
Dual layer architecture	✅ Project+User	❌ Only user level	❌ Only user level	❌ Only at the project level	❌
Multiple IDE sharing	✅	✅	✅	❌	❌
Git version control	✅ Project level traceability	❌	❌	✅	❌
Semantic search	✅	✅	✅	❌	❌
Intelligent classification	✅ Automatic 7-type classification	❌ Manual metadata is required	❌ Manual metadata is required	❌	❌
Sensitive information filtering	✅ Automatically block storage	⚠️ There is PII testing available	⚠️ There is PII testing available	❌	❌
One click installation	✅ `npx omp install`	⚠️ MCP needs to be manually configured	❌ Code integration required	❌	❌
ROT Memory Cleanup	✅ Automatically identify redundancy/obsolescence	❌	❌	❌	❌
Ebbinghaus attenuation	✅ Forgetting Curve Model	❌	❌	❌	❌

Legend: ✅ Complete support| ⚠️ Partial support| ❌ not supported

Key Differences Explanation

Comparison	OpenMemory Plus	Pure OpenMemory MCP
Memory triggered	Automatic extraction at the end of the conversation, without user intervention	Agent needs to actively call `add_memories` Tools
Project context	`_omp/memory/` Storage project decisions and architecture	No project level memory, all memories mixed at user level
Memory classification	Automatically classified as preferences/decisions/patterns, etc.	requires manual input of metadata parameters
Memory Health	ROT testing+health score+decay model	None, memory only increases without decreasing
IDE configuration	Automatically generate AGENTS.md/CLAUDE.md, etc.	Each IDE needs to be manually configured

❓ FAQ

Q: OpenMemory Plus 和 mem0 是什么关系？

A: OpenMemory Plus is mem0enhancement layerNot a substitute.

mem0 Provide underlying vector storage and semantic search capabilities
OpenMemory Plus On this basis, add:
- Dual layer memory architecture (project level+user level)
- Intelligent classification and automatic extraction
- Unified configuration of multiple IDEs
- Sensitive information filtering

You need to install mem0/openmemory MCP first, and then use OpenMemory Plus to enhance it.

Q: 我的数据存储在哪里？安全吗？

A: The data is stored in two places:

Storage location	Data type	Security
`_omp/memory/` (Local)	Project configuration, technical decision-making	✅ Local files, Git version control
`openmemory` (Qdrant)	User preferences, skills	✅ Local Docker container

All data is presentYour local machineup
Sensitive information (API Key, password) may be compromisedauto-filteringNot able to store
Project level data can be obtained through .gitignore exclude

Q: 支持哪些 IDE/Agent？

A: Currently fully supported:

IDE	Configuration Files	Status
Augment	`AGENTS.md`	✅ Complete support
Claude Code	`CLAUDE.md`	✅ Complete support
Cursor	`.cursorrules`	✅ Complete support
Gemini CLI	`gemini.md`	✅ Complete support

Other agents that support MCP protocol can also be used, just manually configure them.

Q: 如何迁移现有项目？

A: Just run the installation command:

cd your-existing-project
npx openmemory-plus install

The installation wizard will:

Detecting existing configurations
create _omp/memory/ catalogue
Generate IDE configuration file
It won't overwrite your existing files

Q: 如何清理过时的记忆？

A: Use the built-in ROT cleaning command:

/mem clean

This will identify and clean:

RRedundant: duplicate information
OObsolete: old information that has been updated
TRivial: a worthless confirmation statement

🗺️ Roadmap

adopt Now / Next / Later Mode, focus on the current direction, and flexibly adjust priorities.

🔥 Now (in progress)

Growth and Community Building

[] Content Marketing - Pain Points Articles, Comparative Videos, Use Cases
[] Community Construction - Discord/WeChat Groups, User Story Collection
[] Document improvement - troubleshooting guide, best practices

Product experience optimization

[] Improved installation success rate - better error prompts and automatic repairs
[] First time experience optimization - feel value within 5 minutes

📅 Next stage

[] Memory Search Enhancement - Semantic Search, Fuzzy Matching
[] Batch Memory Management - CLI commands support batch operations
[] More IDE support - Windsurf, Cline
[] Memory Import/Export - Data Portability

💡 Later (for future consideration)

[] Web UI Management Interface
[] Cloud synchronization options
[] Memory analysis and insights

👉 View the complete Roadmap Issues

✅ Done (v2.0)

已完成功能

v2.0 - Multi-Agent Team Support

[x] 🤝 Multi Agent Team Memory Architecture (Team Shared+Agent Private)
[x] 🎭 Agent ID Schema (actor attribution, scope filtering)
[x] 📋 Team configuration file（_omp/config.yaml)
[x] 🔄 Handoff workflow (task handover protocol between agents)
[x] 📁 Scope Memory API (Double Layer Scope Memory)

V1.6- xMemory Four Layer Architecture

[x] 🧠 XMemory four layer architecture (L3 Theme → L2 Semantic → L1 Episode → L0 Message)
[x] 🎯 L3 theme layer automatic clustering (adsorption/splitting/merging algorithm)
[x] 🔍 Top-down adaptive retrieval（--level theme)
[x] ⚡ Theme search CLI options（--no-expand)
[x] 📊 Connected Component Clustering Algorithm (Implementation of xMemory Paper)

V1.5- Basic Features

[x] Dual layer memory architecture (project level+user level)
[x] Intelligent classification routing
[x] Multiple IDE support (Augment, Claude, cursor, Gemini)
[x] CLI installation tool（npx openmemory-plus install)
[x] Sensitive information filtering
[x] Support for multiple LLM providers (DeepSeek, MiniMax, ZhiPu, Qwen, OpenAI, Ollama)
[x] One click deployment of Docker Compose
[x] Progressive configuration (adding existing files instead of overwriting them)
[x] MCP automatic configuration and verification

🤝 contribution

Welcome to contribute! Please read Contribution Guide Understand how to participate.

# 克隆仓库
git clone https://github.com/Alenryuichi/openmemory-plus.git
cd openmemory-plus

# 安装依赖
cd cli && npm install

# 运行测试
npm test

# 本地开发
npm run dev

📄 license

🔗 Related links

OpenMemory (mem0) - Bottom layer memory service
Qdrant Vector database
BGE-M3 - Multilingual Embedding Model
MCP Protocol — Model Context Protocol

_{Made with ❤️ for AI Agent developers}

使用说明

🎯 Introduction

📸 Effect Display

💡 Why do we need OpenMemory Plus?

Pain point scenario

How to solve OpenMemory Plus?

🎯 Applicable scenarios

👤 Who is suitable for?

📋 Typical use cases

🚀 Quick Start

One-click installation

🐳 Docker Compose One Click Deployment (Recommended)

Basic Usage

System Requirements

✨ feature

core competency

🤖 Multiple LLM support NEW

🧠 Intelligent Memory Management

Advanced features

🏗️ architecture

XMemory four layer memory architecture

Storage Tiering

classification rule

XMemory retrieval process

🖥️ Supported IDEs

📦 CLI commands

Installation command

Supported LLM Providers

🐳 Dependency Service Management (Docker Compose)

Diagnostic command

Agent built-in commands

📁 project structure

🔧 tech stack

⚙️ working principle

Memory retrieval process

Classification Decision Tree

Sensitive information filtering

🔄 Comparison with other schemes

Key Differences Explanation

❓ FAQ

🗺️ Roadmap

🔥 Now (in progress)

📅 Next stage

💡 Later (for future consideration)

✅ Done (v2.0)

🤝 contribution

📄 license

🔗 Related links

相关 MCP

AI持久记忆MCP服务

搜索引擎协议服务器

LLM记忆服务器

YouTube稍后观看清理工具

AI共享记忆数据库

记忆思维管理系统

🤖 Multiple LLM support ^NEW