Multi-Memory MCP Server

Multi-Memory MCP Server

A multi-category knowledge graph memory server using SQLite for persistent storage. Organize memories into isolated contexts for different purposes (work, personal, projects, etc.).

Based on @modelcontextprotocol/server-memory with enhancements:

• SQLite database storage with proper indexing and transactions
• Multi-category support with isolated memory contexts
• LRU connection cache (prevents memory leaks)
• ID-based operations - all objects have unique IDs for precise operations
• Dual identification - use ID or name/type composite key
• Custom properties - JSON properties on entities, observations, and relations (searchable)
• Override mode - update existing records instead of skipping duplicates
• SQL injection protection
• Full test coverage (141 tests)

Quick Start

Run Directly with npx (No Installation Required)

The fastest way to use multi-memory-mcp is to run it directly from GitHub using npx:

npx github:DanNsk/multi-memory-mcp

This will download, build, and run the server automatically. Perfect for trying it out or using in Claude Desktop config:

{
  "mcpServers": {
    "multi-memory": {
      "command": "npx",
      "args": ["github:DanNsk/multi-memory-mcp"],
      "env": {
        "MEMORY_BASE_DIR": "/path/to/.memory",
        "DEFAULT_CATEGORY": "default"
      }
    }
  }
}

Installation (Local Development)

git clone https://github.com/DanNsk/multi-memory-mcp
cd multi-memory-mcp
npm install
npm run build

Configuration

Add to Claude Desktop config:

Config file locations:

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

Using npx (recommended):

{
  "mcpServers": {
    "multi-memory": {
      "command": "npx",
      "args": ["github:DanNsk/multi-memory-mcp"],
      "env": {
        "MEMORY_BASE_DIR": "/Users/yourname/.memory",
        "DEFAULT_CATEGORY": "default"
      }
    }
  }
}

Using local installation (macOS/Linux):

{
  "mcpServers": {
    "multi-memory": {
      "command": "node",
      "args": ["/absolute/path/to/multi-memory-mcp/dist/index.js"],
      "env": {
        "MEMORY_BASE_DIR": "/Users/yourname/.memory",
        "DEFAULT_CATEGORY": "default"
      }
    }
  }
}

Using local installation (Windows):

{
  "mcpServers": {
    "multi-memory": {
      "command": "node",
      "args": ["C:\\path\\to\\multi-memory-mcp\\dist\\index.js"],
      "env": {
        "MEMORY_BASE_DIR": "C:\\Users\\yourname\\.memory",
        "DEFAULT_CATEGORY": "default"
      }
    }
  }
}

Environment Variables

• MEMORY_BASE_DIR: Base directory for all memory categories (default: .aim in current working directory)
• DEFAULT_CATEGORY: Default category when none specified (default: "default")
• SERIALIZATION_FORMAT: Output format for tool responses (default: "json")
  • json - Standard JSON with 2-space indentation
  • toon - TOON (Token-Oriented Object Notation) - compact format optimized for LLMs with 30-60% fewer tokens

TOON Format

When SERIALIZATION_FORMAT=toon, responses use TOON format which is more token-efficient for LLM contexts.

Structure:

• Objects: key: value with 2-space indentation for nesting
• Arrays: name[count]{field1,field2}: followed by comma-separated rows
• Primitives: unquoted unless containing special characters

Escaping rules (only these escape sequences are valid):

• \\ - backslash
• \" - double quote
• \n - newline
• \r - carriage return
• \t - tab

Quoting required when: empty string, leading/trailing spaces, matches true/false/null, numeric, or contains : " \ [ ] { } ,

Example JSON vs TOON:

JSON (standard):

{
  "entities": [
    {"id": "1", "name": "AuthService", "entityType": "module", "observations": []}
  ]
}

TOON (compact):

entities[1]{id,name,entityType,observations}:
  1,AuthService,module,[]

See TOON specification for full format details.

Database Schema

Each category stores data in a separate SQLite database with the following schema:

Tables

`entities`

Primary storage for graph nodes.

Unique Constraint: (name, entity_type) - entities are identified by name+type combination

`observations`

Facts and notes associated with entities.