/ 目录 / 演练场 / claude-mem
← 全部服务器 ↗ GitHub
● 社区 thedotmack ⚡ 即开即用

claude-mem

作者 thedotmack · thedotmack/claude-mem

claude-mem captures every Claude Code session, compresses it with AI into a searchable memory, and injects only the relevant bits into future sessions.

claude-mem is a plugin/skill hybrid for Claude Code (also Gemini CLI and OpenCode). Five lifecycle hooks observe everything your agent does; an AI compression pass distills sessions into structured observations; a SQLite + Chroma vector store powers semantic recall. You search memory with the mem-search skill, and a local web viewer at localhost:37777 shows the memory stream in real time. Privacy controls via <private> tags keep secrets out. The biggest win: no more 'remind me what we decided last Tuesday' at the start of every session.

▶ 观看演示 📦 安装 💡 使用场景

为什么要用

核心特性

实时演示

实际使用效果

claude-mem-skill.replay ▶ 就绪
0/0

安装

选择你的客户端

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "claude-mem-skill": {
      "command": "npx",
      "args": [
        "claude-mem",
        "install"
      ]
    }
  }
}

打开 Claude Desktop → Settings → Developer → Edit Config。保存后重启应用。

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "claude-mem-skill": {
      "command": "npx",
      "args": [
        "claude-mem",
        "install"
      ]
    }
  }
}

Cursor 使用与 Claude Desktop 相同的 mcpServers 格式。项目级配置优先于全局。

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "claude-mem-skill": {
      "command": "npx",
      "args": [
        "claude-mem",
        "install"
      ]
    }
  }
}

点击 Cline 侧栏中的 MCP Servers 图标，然后选 "Edit Configuration"。

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "claude-mem-skill": {
      "command": "npx",
      "args": [
        "claude-mem",
        "install"
      ]
    }
  }
}

格式与 Claude Desktop 相同。重启 Windsurf 生效。

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "claude-mem-skill",
      "command": "npx",
      "args": [
        "claude-mem",
        "install"
      ]
    }
  ]
}

Continue 使用服务器对象数组，而非映射。

~/.config/zed/settings.json
{
  "context_servers": {
    "claude-mem-skill": {
      "command": {
        "path": "npx",
        "args": [
          "claude-mem",
          "install"
        ]
      }
    }
  }
}

加入 context_servers。Zed 保存后热重载。

claude mcp add claude-mem-skill -- npx claude-mem install

一行命令搞定。用 claude mcp list 验证，claude mcp remove 卸载。

使用场景

实战用法： claude-mem

Resume a multi-week project without re-briefing Claude every morning

👤 Engineers on long-running features or research ⏱ ~5 min beginner

何时使用： You spend the first 10 minutes of every session re-explaining context that Claude had yesterday.

前置条件
  • Claude Code installed — npm install -g @anthropic-ai/claude-code
  • claude-mem installed — npx claude-mem install (registers hooks in ~/.claude)
步骤
  1. Work normally
    Just use Claude Code. claude-mem is passive — hooks capture events automatically.✓ 已复制
    → localhost:37777 shows the memory stream updating as you work
  2. Start tomorrow's session
    Where did we leave off on the migration branch?✓ 已复制
    → Claude recalls concrete facts + cites observation IDs
  3. Correct misremembering
    That's wrong about the DB schema — we decided to keep the v1 field. Mark that observation as corrected.✓ 已复制
    → mem-search writes a correction; future recall prioritizes it

结果： Sessions feel continuous. The re-briefing ritual disappears.

注意事项
  • Memory recalls stale facts after you change course — Explicitly mark corrections; mem-search respects recency and overrides
  • Sensitive code snippets get captured — Wrap them in <private>...</private> tags; hooks skip private blocks
搭配使用： filesystem

Turn individual sessions into team knowledge without writing docs

👤 Team leads; ICs who avoid writing postmortems ⏱ ~10 min intermediate

何时使用： You solve a gnarly bug, explain the fix to Claude, and want that explanation findable next month without writing a README.

步骤
  1. Narrate your fix
    Walk me through why the deadlock happened and the fix you just landed. Be complete — future me might read this.✓ 已复制
    → Claude writes a narrative; compression captures it
  2. Tag the memory
    Tag this observation with 'db-deadlock-2026-04' and 'billing-cluster'.✓ 已复制
    → Tags stored on the observation
  3. Recall later
    A month later: 'We've got another deadlock in billing — pull any prior notes.'✓ 已复制
    → The April note surfaces first

结果： Lightweight knowledge base that grows by itself.

注意事项
  • Memory gets noisy as months pass — Use mem-search --prune on low-signal old sessions; Endless Mode (beta) handles this automatically
搭配使用： memory

Keep customer data and secrets out of memory captures

👤 Anyone working with sensitive data ⏱ ~15 min intermediate

何时使用： You work on code that touches PII or secrets and can't let them land in a local vector store.

步骤
  1. Wrap sensitive blocks
    Write me the sample data with real test records wrapped in <private> tags so claude-mem doesn't capture the values.✓ 已复制
    → Private blocks are in the transcript but not in the memory store
  2. Verify
    Open localhost:37777 and check today's stream. Confirm no records from the <private> block appear.✓ 已复制
    → Private content absent
  3. Add repo-wide policy
    Update CLAUDE.md: 'Always wrap real customer data in <private>'.✓ 已复制
    → CLAUDE.md updated; future sessions inherit

结果： Memory is useful without being a data-leak surface.

注意事项
  • Forgetting to tag sensitive content — Make it a pre-commit / hook convention; claude-mem can enforce prompts that remind on detection

组合

与其他 MCP 搭配，撬动十倍杠杆

claude-mem-skill + filesystem

Memory recalls the file you edited, filesystem re-reads it

Pull from mem-search the last file we edited in the billing refactor; then read it and continue where we left off.✓ 已复制
claude-mem-skill + serena

Memory supplies the intent; Serena supplies the current code state

Memory says we're renaming ChargeService → BillingService; use Serena to finish the rename.✓ 已复制
claude-mem-skill + github

Cross-reference memory with real PR state

Memory says we opened a PR yesterday; use GitHub MCP to get its current status and review comments.✓ 已复制

工具

此 MCP 暴露的能力

工具输入参数何时调用成本
mem-search (skill) query: str, limit?: int, since?: date Any time you'd ask Claude to 'remember' — full-text and semantic search over your history free (local SQLite + embedding model)
mem-compress (hook) session transcript Automatic — runs at session end; rarely invoked manually small Claude call per session
web viewer URL http://localhost:37777 Debug what's being captured; check privacy filters free

成本与限制

运行它的成本

API 配额
Local SQLite + Chroma vectors; embedding model bundled
每次调用 Token 数
Small — observations are already compressed
费用
Free (MIT)
提示
Compression calls use your Claude API key; set a monthly budget in the claude-mem config if you chat a lot

安全

权限、密钥、影响范围

凭据存储： None for the tool itself; compression uses your existing Claude API key
数据出站： Compression calls go to Anthropic (your account); nothing else leaves your machine
切勿授予： Uploading the SQLite/Chroma store off-device without review — it contains transcripts

故障排查

常见错误与修复

Web viewer at localhost:37777 won't load

Check the hook is running: ps aux | grep claude-mem. Reinstall with npx claude-mem install if it's absent.

验证： curl -I http://localhost:37777
Recall returns nothing useful

First few sessions are cold; memory gets better after ~10 sessions as patterns emerge. Also check that hooks are actually firing (see viewer).

验证： Open viewer during a session; events should stream in real time
Compression calls are expensive

Lower compression depth in config or use Endless Mode (beta) which is cheaper per session

验证： Check Anthropic usage dashboard for attribution
<private> content still shows up

Tag must be on its own lines; inline <private>…</private> in the middle of a sentence isn't parsed. Use block form.

验证： Reproduce and inspect the stored observation in the viewer

替代方案

claude-mem 对比其他方案

替代方案何时用它替代权衡
Memory MCPYou want explicit, user-controlled memory writes — not passive captureYou do the memory-saving yourself; less ambient, less magic
Serena memoryYour use case is code-focused and you'd rather have per-project notes than transcript compressionNo session-stream capture; more project-level
Just take notesYou prefer explicit, human-curated notesNo automation; discipline-dependent

更多

资源

📖 阅读 GitHub 上的官方 README

🐙 查看未解决的 issue

🔍 浏览全部 400+ MCP 服务器和 Skills