# nemp-mcp-server

Persistent local memory for AI agents. MCP server that works with Claude Code, Cursor, Windsurf, Emergent, and any MCP-compatible client.

## Quick Start

Add this to your MCP client config:

{

  "nemp-memory": {

    "command": "npx",

    "args": \["-y", "nemp-mcp-server"]

  }

}

Restart your MCP client.

## Tools

- nemp_init — Auto-detect project stack and save as memories

- nemp_save — Save a key-value memory with compression

- nemp_recall — Search memories with semantic keyword expansion

- nemp_list — List all stored memories

- nemp_forget — Delete a memory by key

- nemp_context — Return full project context for agent injection

- nemp_log — View access audit trail

## Features

- Auto-detects your tech stack from package.json, requirements.txt, etc.

- Persists decisions, conventions, and context across sessions

- Shares memory between multiple agents on the same project

- Tracks which agent wrote each memory with audit logs

- Compresses values by ~75% to save context window

- 100% local — zero cloud, zero network calls, plain JSON

## How It Works

Memories are stored in .nemp/ in your project root as JSON files. Every MCP client connected to the same project shares the same memory.

## Links

- Website: https://nemp.dev

- GitHub: https://github.com/SukinShetty/Nemp-memory

- npm: https://www.npmjs.com/package/nemp-mcp-server