shared-memory-mcp
跨 IDE 共享记忆的 MCP Server。支持 Kiro、Cursor、Windsurf、Trae 等多终端间的会话记忆共享。
Agent 自动判断写入时机,无需人工干预。使用 SQLite + FTS5,零外部依赖。
安装
# 方式一:uvx 直接运行(推荐,无需安装)
uvx shared-memory-mcp
# 方式二:pip 安装
pip install shared-memory-mcp
配置 MCP
方式一:uvx(推荐)
发布到 PyPI 后,在各 IDE 的 MCP 配置中添加:
{
"mcpServers": {
"shared-memory": {
"command": "uvx",
"args": ["shared-memory-mcp"],
"disabled": false,
"autoApprove": ["memory_write", "memory_read", "memory_list", "memory_delete"]
}
}
}
方式二:本地源码运行
{
"mcpServers": {
"shared-memory": {
"command": "python",
"args": ["-m", "shared_memory_mcp.server"],
"env": {
"PYTHONPATH": "D:\\github\\mcp-sharedMemory\\shared-memory-mcp\\src"
},
"disabled": false,
"autoApprove": ["memory_write", "memory_read", "memory_list", "memory_delete"]
}
}
}
各 IDE 配置文件位置
| IDE | 项目级配置 | 全局配置 |
|---|---|---|
| Kiro | .kiro/settings/mcp.json |
~/.kiro/settings/mcp.json |
| Cursor | .cursor/mcp.json |
~/.cursor/mcp.json |
| Windsurf | 对应 MCP 配置位置 | 对应全局配置 |
| Trae | 对应 MCP 配置位置 | 对应全局配置 |
配置 Steering 规则
在各 IDE 的全局规则中添加以下内容,让 Agent 自动读写记忆:
## 共享记忆规则
### 项目标识
- project 参数使用「父级文件夹/当前文件夹」拼接,如果父级是磁盘根目录则只用当前文件夹名
- 例:D:\github\my-app → project="github/my-app"
- 例:D:\my-app → project="my-app"
### 会话开始
- 自动调用 memory_list(project="<当前项目名>") 获取记忆索引
- 如有相关记忆,调用 memory_read 获取最近关键记忆的摘要
### 自动写入时机
当你做出以下行为时,自动调用 memory_write 写入共享记忆:
- 架构决策(type="decision")
- 阶段完成(type="milestone")
- 重要结论(type="conclusion")
- 项目上下文(type="context")
- 问题方案(type="issue")
### 用户手动触发
- "记住:..." / "记录:..." — 直接写入
- "记录上面的关键点" / "总结记录一下" — 回顾对话提取关键内容写入
- "记录上一轮对话" — 仅回顾上一个来回
### 不记录
- 简单问答、闲聊、临时调试等低价值信息
各 IDE 规则文件位置
| IDE | 全局规则路径 |
|---|---|
| Kiro | ~/.kiro/steering/shared-memory.md |
| Cursor | Cursor Settings → Rules → User Rules |
工具说明
| 工具 | 说明 | Token 消耗 |
|---|---|---|
memory_write |
写入记忆,返回仅 id+status | 极低 |
memory_read |
读取记忆,默认只返回摘要 | 低 |
memory_list |
极简索引,每条一行 | 最低 |
memory_delete |
删除过时记忆 | 极低 |
发布到 PyPI
前置条件
- 注册 PyPI 账号
- 安装构建工具:
pip install build twine
构建与发布
# 进入项目目录
cd shared-memory-mcp
# 构建
python -m build
# 上传到 TestPyPI(先测试)
twine upload --repository testpypi dist/*
# 从 TestPyPI 测试安装
pip install --index-url https://test.pypi.org/simple/ shared-memory-mcp
# 确认无误后,上传到正式 PyPI
twine upload dist/*
发布后验证
# uvx 直接运行(无需安装)
uvx shared-memory-mcp
# 或 pip 安装后运行
pip install shared-memory-mcp
shared-memory-mcp
数据存储
- 数据库文件:
~/.shared-memory-mcp/memories.db - 所有 IDE 共享同一个 SQLite 数据库
- 使用 WAL 模式支持并发读取
- FTS5 全文搜索索引
License
MIT
Metadata
Release files for shared-memory-mcp 0.4.0
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | Provenance | |
|---|---|---|---|---|
| shared_memory_mcp-0.4.0.tar.gz | 8.0 kB | Without provenance | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| shared_memory_mcp-0.4.0-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 17.3 kB
Security for shared-memory-mcp 0.4.0
Provenance
Provenance describes where a file came from. On PyPI, provenance is shared via attestations, which provide a verifiable record of the build or publishing details. View details, limitations and caveats.
No provenance found
None of the files in this release were uploaded with provenance metadata.
A note on package security
PyPI is a repository for distributing packages, not a service for auditing them. While third-party security researchers and our staff do their best to detect and remove malicious packages, PyPI cannot guarantee the quality, security or suitability of any given package.
To learn more about protecting yourself from malicious actors, visit our package security documentation.