sourknives/memory-mcp-server

Cortex MCP

A Model Context Protocol (MCP) server providing intelligent, persistent memory storage across AI development tools.

Overview

Cortex MCP enables AI tools like Claude Desktop, Cursor, and Kiro to share persistent memory and context across conversations and sessions.

Key Features:

• 🧠 Persistent Memory - Store and retrieve information across AI tool sessions
• 🔍 Semantic Search - Find relevant information using AI-powered similarity
• 🔒 Local & Private - Runs on your infrastructure with encrypted storage
• 🌐 Multi-Tool Support - Works with Claude Desktop, Cursor, Kiro, and other MCP-enabled tools
• 📊 Intelligent Storage - Automatically categorizes and stores valuable conversations

Quick Start

For Remote Server Usage (Recommended)

1. Install the lightweight client:

   pip install cortex-mcp-client
   

2. Configure your AI tool (Claude Desktop, Cursor, Kiro):

   {
     "mcpServers": {
       "cortex-mcp": {
         "command": "cortex-mcp-client",
         "args": [],
         "env": {
           "CORTEX_MCP_SERVER_URL": "https://your-server.com:8000",
           "CORTEX_MCP_API_KEY": "your-api-key"
         }
       }
     }
   }
   

3. Test the connection:

   cortex-mcp-client --test
   

For Local Development

1. Set up the server:

   cd cortex-mcp-server
   pip install -e .
   python -m cortex_mcp.main
   

2. Configure your AI tool for local connection:

   {
     "mcpServers": {
       "cortex-mcp": {
         "command": "python",
         "args": ["-m", "cortex_mcp.main", "--mode", "mcp"],
         "cwd": "/path/to/cortex-mcp-server"
       }
     }
   }
   

Repository Structure

cortex-mcp/
├── cortex-mcp-server/     # Full MCP server implementation
│   ├── src/               # Server source code
│   ├── docs/              # Server documentation
│   ├── scripts/           # Management scripts
│   └── tests/             # Server tests
├── cortex-mcp-client/     # Lightweight remote client
│   ├── client.py          # Core MCP-to-HTTP bridge
│   ├── main.py            # CLI interface
│   └── README.md          # Client documentation
└── README.md              # This file

Available Tools

Once configured, these MCP tools are available in your AI development environment:

• store_memory - Store information in persistent memory
• retrieve_memories - Retrieve stored memories by query
• search_memories - Semantic search across all memories
• list_memories - List recent or filtered memories
• delete_memory - Remove specific memories
• analyze_conversation_for_storage - Analyze conversation value
• suggest_memory_storage - Get intelligent storage suggestions

Configuration Locations

Claude Desktop

• macOS: ~/Library/Application Support/Claude/claude_desktop_config.json
• Windows: %APPDATA%\Claude\claude_desktop_config.json

Kiro IDE

• Workspace: .kiro/settings/mcp.json
• User: ~/.kiro/settings/mcp.json

Cursor IDE

• Configuration location varies by platform

Documentation

• Server Setup: See cortex-mcp-server/README.md
• Client Usage: See cortex-mcp-client/README.md
• Deployment: See cortex-mcp-server/DEPLOYMENT.md

Support

• Issues: GitHub Issues
• Documentation: GitHub Repository

License

MIT License - see LICENSE file for details.