Content
# Personal Memory MCP Server
> A lightweight, seamless, and concise personal AI memory engine — your dedicated intelligent assistant that remembers you across every conversation.
## What is this
Personal Memory is an [MCP](https://modelcontextprotocol.io/) (Model Context Protocol) Server that gives Claude, Cursor, and other AI hosts **persistent personal memory**.
Once installed, your AI remembers your habits, preferences, experiences, and relationships — no more starting from scratch every conversation.
## Features
| Tool | Description |
|------|-------------|
| `add_memory` | Record events, decisions, insights (auto-deduplicated) |
| `search_memories` | Full-text search across all memories (FTS5) |
| `get_memory_summary` | Get a complete personal profile overview |
| `get_daily_briefing` | Daily briefing: yesterday's memories + follow-ups + habit alerts |
| `get_follow_ups` | Pending plans, unactioned decisions, lost contacts |
| `check_habit_patterns` | Detect habit anomalies (e.g., no exercise for 3 days) |
| `upsert_person` | Record / update a person and their relationship |
| `feed_habit` | Record a user habit |
| `feed_preference` | Record a user preference |
| `import_wechat` | Import WeChat chat logs, auto-extract events |
| `import_browser_history` | Read Chrome/Edge history, extract interests |
| `archive_old_memories` | Archive old low-importance memories |
## Quick Start
### Option 1: One-line install (recommended)
**Windows (PowerShell):**
```powershell
irm https://raw.githubusercontent.com/Emama-bit/personal-memory-mcp/master/install.ps1 | iex
```
**macOS / Linux:**
```bash
curl -fsSL https://raw.githubusercontent.com/Emama-bit/personal-memory-mcp/master/install.sh | bash
```
### Option 2: Manual install
**Windows (PowerShell):**
```powershell
irm https://raw.githubusercontent.com/Emama-bit/personal-memory-mcp/master/install.ps1 | iex
```
**macOS / Linux:**
```bash
curl -fsSL https://raw.githubusercontent.com/Emama-bit/personal-memory-mcp/master/install.sh | bash
```
Restart Claude Desktop / Cursor after installation.
### Option 2: Manual install
```bash
git clone https://github.com/Emama-bit/personal-memory-mcp.git
cd personal-memory-mcp/agent
npm install
```
Then configure your AI host:
**Claude Desktop** — Settings → Developer → Edit Config:
```json
{
"mcpServers": {
"personal-memory": {
"command": "npx",
"args": ["tsx", "/path/to/personal-memory-mcp/agent/src/server.ts"]
}
}
}
```
**Cursor** — Create `.cursor/mcp.json`:
```json
{
"mcpServers": {
"personal-memory": {
"command": "npx",
"args": ["tsx", "/path/to/personal-memory-mcp/agent/src/server.ts"]
}
}
}
```
Restart your AI host after configuring.
## Usage Example
```
You: My name is Alex, I wake up at 7am every day and prefer VS Code
AI: [Silently calls feed_habit + feed_preference]
You: Do you remember me?
AI: Of course! You're Alex, up at 7am, coding in VS Code...
You: Search my memories about Rust
AI: You decided to use Rust for your new project because...
```
## Architecture
```
Claude / Cursor / Any MCP Host
│ MCP Protocol (stdio)
▼
Personal Memory Server
├── Memory Engine (SQLite + FTS5)
├── People / Relationship Tracking
├── Personal Profile (habits, preferences)
├── Proactive Intelligence (daily briefing, follow-ups)
└── Data Imports (WeChat, browser history)
```
- **Storage**: SQLite with WAL mode, atomic writes
- **Search**: FTS5 full-text index, supports Chinese & English
- **Dedup**: Auto-detect similar memories before writing
- **Tiered**: 7-day / 30-day / archived layers, system prompt stays compact
- **Privacy**: All data stored locally, never uploaded
## Project Structure
```
agent/
├── src/
│ ├── server.ts ← MCP Server entry (18 tools)
│ ├── memory/
│ │ ├── db.ts ← SQLite init + FTS5
│ │ ├── memories.ts ← Memory CRUD + search + dedup
│ │ ├── people.ts ← People / relationship management
│ │ └── conversations.ts ← Conversation persistence
│ ├── profile.ts ← User profile (habits, preferences)
│ ├── system.ts ← Tiered system prompt generation
│ ├── wechat.ts ← WeChat chat log parser
│ └── browser.ts ← Chrome/Edge history reader
├── memory/agent.db ← Local SQLite database (gitignored)
└── package.json
```
## Contact
If you have any questions or suggestions, please contact me: lk2012943494@outlook.com
## License
MIT
