🧠 Memory MCP Server

> Persistent memory and context window caching for LLM conversations. Save, retrieve, and manage memories with intelligent context archiving. MongoDB-backed storage.

An MCP (Model Context Protocol) server that provides memory management and context window caching for AI coding environments like Cursor and Claude Desktop.

Why Use Memory MCP?

• 💾 Persistent Storage – MongoDB-backed memory that survives sessions
• 🧠 Context Caching – Intelligent archiving and retrieval of conversation context
• 🏷️ Tag-based Search – Organize and find memories by tags
• 📊 Relevance Scoring – Automatically score archived content relevance
• ⚡ Easy Setup – One-click install in Cursor or simple manual setup

Quick Start

Ready to add memory to your AI workflow? Install in seconds:

Install in Cursor (Recommended):

🔗 Install in Cursor

Or install manually:

npm install -g @jamesanz/memory-mcp
# Or from source:
git clone https://github.com/JamesANZ/memory-mcp.git
cd memory-mcp && npm install && npm run build

Features

Basic Memory Tools

• `save-memories` – Save memories to database (overwrites existing)
• `get-memories` – Retrieve all stored memories
• `add-memories` – Append new memories without overwriting
• `clear-memories` – Remove all stored memories

Context Window Caching

• `archive-context` – Archive conversation context with tags
• `retrieve-context` – Retrieve relevant archived context
• `score-relevance` – Score archived content relevance
• `create-summary` – Create summaries of archived content
• `get-conversation-summaries` – Get all summaries for a conversation
• `search-context-by-tags` – Search archived content by tags

Installation

Cursor (One-Click)

Click the install link above or use:

cursor://anysphere.cursor-deeplink/mcp/install?name=memory-mcp&config=eyJtZW1vcnktbWNwIjp7ImNvbW1hbmQiOiJucHgiLCJhcmdzIjpbIi15IiwiQGphbWVzYW56L21lbW9yeS1tY3AiXX19

Manual Installation

Requirements: Node.js 18+, npm, MongoDB

# Clone and build
git clone https://github.com/JamesANZ/memory-mcp.git
cd memory-mcp
npm install
npm run build

# Set MongoDB connection string
export MONGODB_URI="mongodb://localhost:27017"

# Run server
npm start

Claude Desktop

Add to `claude_desktop_config.json`:

macOS: `~/Library/Application Support/Claude/claude_desktop_config.json`

Windows: `%APPDATA%\Claude\claude_desktop_config.json`

{
  "mcpServers": {
    "memory-mcp": {
      "command": "node",
      "args": ["/absolute/path/to/memory-mcp/build/index.js"],
      "env": {
        "MONGODB_URI": "mongodb://localhost:27017"
      }
    }
  }
}

Restart Claude Desktop after configuration.

Configuration

Set the MongoDB connection string:

export MONGODB_URI="mongodb://localhost:27017"

Default: `mongodb://localhost:27017`

Usage Examples

Save Memories

Store memories from a conversation:

{
  "tool": "save-memories",
  "arguments": {
    "memories": ["User prefers TypeScript", "User works on blockchain projects"],
    "llm": "claude",
    "userId": "user123"
  }
}

Retrieve Memories

Get all stored memories:

{
  "tool": "get-memories",
  "arguments": {}
}

Archive Context

Archive conversation context when it gets too long:

{
  "tool": "archive-context",
  "arguments": {
    "conversationId": "conv-123",
    "contextMessages": ["Message 1", "Message 2"],
    "tags": ["coding", "typescript"],
    "llm": "claude"
  }
}

Retrieve Relevant Context

Get archived content relevant to current conversation:

{
  "tool": "retrieve-context",
  "arguments": {
    "conversationId": "conv-123",
    "tags": ["coding"],
    "minRelevanceScore": 0.5,
    "limit": 10
  }
}

Context Window Caching

The system automatically:

• Archives content when context usage reaches 80%
• Retrieves relevant content when usage drops below 30%
• Scores relevance using keyword overlap
• Creates summaries to condense long conversations

Use Cases

• Long Conversations – Manage context windows for extended sessions
• Memory Persistence – Save important information across sessions
• Context Retrieval – Bring back relevant past conversations
• Research Projects – Organize and tag research conversations

Technical Details

Built with: Node.js, TypeScript, MCP SDK, MongoDB

Dependencies: `@modelcontextprotocol/sdk`, `mongodb`, `zod`

Platforms: macOS, Windows, Linux

Storage: MongoDB (default: `mongodb://localhost:27017`)

Contributing

⭐ If this project helps you, please star it on GitHub! ⭐

Contributions welcome! Please open an issue or submit a pull request.

License

ISC

Support

If you find this project useful, consider supporting it:

⚡ Lightning Network

lnbc1pjhhsqepp5mjgwnvg0z53shm22hfe9us289lnaqkwv8rn2s0rtekg5vvj56xnqdqqcqzzsxqyz5vqsp5gu6vh9hyp94c7t3tkpqrp2r059t4vrw7ps78a4n0a2u52678c7yq9qyyssq7zcferywka50wcy75skjfrdrk930cuyx24rg55cwfuzxs49rc9c53mpz6zug5y2544pt8y9jflnq0ltlha26ed846jh0y7n4gm8jd3qqaautqa

₿ Bitcoin: bc1ptzvr93pn959xq4et6sqzpfnkk2args22ewv5u2th4ps7hshfaqrshe0xtp

Ξ Ethereum/EVM: 0x42ea529282DDE0AA87B42d9E83316eb23FE62c3f