Tech Memory Mcp — independently scanned and version-tracked by SaferSkills.
SaferSkills independently audited Tech Memory Mcp (Agent Skill) and scored it 100/100 (green). The audit ran 55 deterministic rules across Security, Supply Chain, Maintenance, Transparency, and Community; it found 0 high-severity and 0 lower-severity findings. The full rule-by-rule trace and per-finding evidence are below. Free, methodology-open.
Findings & checks · 0 flagged
Every scanned point with the score it earned and what moved between them.
First recorded scan — no prior version to compare against.
The primary manifest — the file an agent reads to learn what this artifact does.
不只是记住"说了什么",而是记住"学到了什么"。
让 Claude Code 等 AI 编程助手具备长期记忆能力,从每次技术对话中自动提取知识点,下次对话时自动检索相关经验。
npm install --global windows-build-tools#### 方式一:npx 安装(推荐)
npx -y tech-memory-mcp@latest首次运行会自动下载嵌入模型(约 154MB,30-60 秒)。模型仅需下载一次,后续启动直接使用缓存。
#### 方式二:从源码安装
git clone https://github.com/myx0423/tech-memory-mcp.git
cd tech-memory-mcp
npm install
npm run build然后将 .mcp.json 配置改为:
{
"mcpServers": {
"tech-memory": {
"command": "node",
"args": ["/你的路径/tech-memory-mcp/dist/index.js"]
}
}
}在项目根目录或 ~/.claude/ 创建 .mcp.json:
{
"mcpServers": {
"tech-memory": {
"command": "npx",
"args": ["-y", "tech-memory-mcp"]
}
}
}启动 Claude Code,输入:
请调用 tech_stats 查看知识库统计如果返回知识库统计信息(即使为空),说明安装成功。首次启动会自动下载模型,约 30-60 秒。
如果网络环境不佳,可以提前手动下载模型:
从 HuggingFace 镜像站下载模型到本地目录:
mkdir -p ~/models/jina-embeddings-v2-base-zh
cd ~/models/jina-embeddings-v2-base-zh
# 使用 wget 下载(或使用浏览器手动下载)
wget https://hf-mirror.com/Xenova/jina-embeddings-v2-base-zh/resolve/main/config.json
wget https://hf-mirror.com/Xenova/jina-embeddings-v2-base-zh/resolve/main/model.onnx
wget https://hf-mirror.com/Xenova/jina-embeddings-v2-base-zh/resolve/main/tokenizer.json
wget https://hf-mirror.com/Xenova/jina-embeddings-v2-base-zh/resolve/main/vocab.txt修改 .mcp.json,添加 TECH_MEMORY_MODEL_PATH:
{
"mcpServers": {
"tech-memory": {
"command": "npx",
"args": ["-y", "tech-memory-mcp"],
"env": {
"TECH_MEMORY_MODEL_PATH": "/Users/yourname/models/jina-embeddings-v2-base-zh"
}
}
}
}Windows 用户示例:
{
"mcpServers": {
"tech-memory": {
"command": "npx",
"args": ["-y", "tech-memory-mcp"],
"env": {
"TECH_MEMORY_MODEL_PATH": "C:\\Users\\yourname\\models\\jina-embeddings-v2-base-zh"
}
}
}
}可选:自定义模型缓存目录(用于自动下载的模型):
{
"mcpServers": {
"tech-memory": {
"command": "npx",
"args": ["-y", "tech-memory-mcp"],
"env": {
"TECH_MEMORY_CACHE_DIR": "/path/to/cache"
}
}
}
}| 工具 | 作用 |
|---|---|
tech_search | 混合搜索(语义 + 全文),支持置信度过滤和过期知识过滤,自动记录曝光事件 |
tech_store | 存储知识点,自动去重,支持置信度和过期时间 |
tech_extract_template | 返回中文知识提取 Prompt + JSON Schema |
tech_auto_extract | 自动从对话中提取并存储技术知识(支持两阶段调用) |
tech_confirm | 确认知识点是否有用,调整置信度,记录采用/拒绝事件 |
tech_decay | 执行知识老化衰减,批量更新置信度,清理过期使用事件 |
tech_outdated | 查询过期或低置信度的知识点 |
tech_conflict_scan | 扫描知识库中的冲突和重复条目 |
tech_resolve | 处理冲突条目(保留、合并或标记矛盾) |
tech_link | 创建知识点之间有向关系 |
tech_get | 查询知识点 + 关联条目 |
tech_stats | 数据库统计 |
tech_usage_stats | 知识库使用统计(曝光、采用、采用率、热门查询) |
tech_export | 导出知识库为 Markdown 或 JSON 格式 |
tech_import | 从 JSON/Markdown/纯文本批量导入知识 |
tech_backup | 创建 SQLite 数据库完整备份 |
| 类型 | 说明 | 示例 |
|---|---|---|
decision | 技术决策 | "选了 Calico 而非 Flannel,因为需要 NetworkPolicy" |
lesson | 经验教训 | "kubeadm init 失败记得先 kubeadm reset -f" |
preference | 个人偏好 | "更喜欢用 Containerd 而非 Docker 作为运行时" |
fact | 技术事实 | "K8s v1.34 对应 Containerd 沙箱镜像 pause:3.10" |
pattern | 通用模式 | "三 Master + Nginx + KeepAlived 是生产标配" |
对话结束 → tech_auto_extract 提取知识点
→ tech_store 存储(自动嵌入 + 语义去重)
→ 下次对话时 tech_search 自动检索相关记忆
→ tech_confirm 反馈知识价值,优化搜索排序~/.tech-memory/memory.db(SQLite + 内置向量搜索)Xenova/jina-embeddings-v2-base-zh(q8 量化,154 MB)git clone https://github.com/myx0423/tech-memory-mcp.git
cd tech-memory-mcp
npm install
npm run build
node dist/index.js # 启动 MCP Server运行测试:
npm test| 项目 | 说明 |
|---|---|
| remote-ops-mcp | SSH 远程服务器管理:执行命令、部署服务、排查故障。操作失败时自动缓存踩坑上下文,配合 tech_memory.tech_store 形成运维知识闭环。 |
MIT
~30 seconds. Free. No account. Every finding cites a rule and a line of evidence.