Memory MCP

What it does

Memory MCP gives AI assistants persistent, searchable memory that survives across sessions and LLM surfaces. It runs as an MCP server backed by a local SQLite database and renders an interactive dashboard panel directly inside Claude Desktop and VS Code Copilot — no browser required. You can save preferences, search past context, and bulk-import conversation history from ChatGPT exports, Gemini Takeout, or Markdown notes.

Key features

• Interactive Dashboard UI — rendered as a panel inside Claude Desktop or VS Code Copilot via MCP Apps; browse, filter, and delete memories without leaving your AI client
• Import Wizard — step-by-step guided import from ChatGPT JSON exports, Gemini Takeout, .cursorrules, or plain Markdown
• Full-text search — FTS5-powered search with AND/OR/NOT operators and quoted phrases
• Source-tagged memories — each memory tracks which LLM surface it came from (claude, chatgpt, gemini, cursor, manual)
• CLI (mem) — add, search, list, delete, and import memories from the terminal independently of any AI client

Installation

Claude Desktop

Edit ~/Library/Application Support/Claude/claude_desktop_config.json:

{
  "mcpServers": {
    "memory-mcp": {
      "command": "node",
      "args": ["/path/to/memory-mcp/build/index.js"]
    }
  }
}

Then ask Claude: "Show me my memory dashboard" to open the UI panel.

VS Code Copilot

Add to settings.json:

{
  "github.copilot.chat.mcp.servers": {
    "memory-mcp": {
      "command": "node",
      "args": ["/path/to/memory-mcp/build/index.js"]
    }
  }
}

Cursor

Add to ~/.cursor/mcp.json:

{
  "memory-mcp": {
    "command": "node",
    "args": ["/path/to/memory-mcp/build/index.js"]
  }
}

Supported hosts

Confirmed: Claude Desktop, Claude Code, VS Code Copilot, Cursor, Windsurf.