Memory Decay

A Model Context Protocol (MCP) server that implements memory with decay mechanics. Memories fade over time unless accessed, mimicking human memory retention. It also includes a permanent "journal" for verification.

Why This?

Inspired by https://www.moltbook.com/post/783de11a-2937-4ab2-a23e-4227360b126f

Prerequisites

• uv (Fast Python package installer and resolver)
• Python 3.12+

Setup

1. Clone the repository:

   git clone <repository-url>
   cd PyMemoryDecayMCP
   

2. Install dependencies: Initialize the project and install dependencies using uv:

   uv sync
   

Running Locally

To run the MCP server from source:

uv run main.py

Building the Executable

To compile the project into a single standalone executable (.exe):

1. Add development dependencies (if not already added):

   uv add --dev pyinstaller
   

2. Build:

   uv run pyinstaller --onefile main.py --name memory-decay-mcp --clean
   

3. Locate the executable: The compiled .exe will be found in the dist/ directory: dist/memory-decay-mcp.exe

VS Code Configuration

To use this with the VS Code MCP extension, add the following to your mcp.json config file (typically found in %APPDATA%\Code\User\globalStorage\mcp-server\mcp.json or configured via the extension settings).

Using Source (Recommended for Dev)

{
  "mcpServers": {
    "memory-decay": {
      "command": "uv",
      "args": [
        "run",
        "C:\\path\\to\\PyMemoryDecayMCP\\main.py"
      ],
      "env": {
        "MEMORY_STORAGE_PATH": "C:\\path\\to\\custom\\data\\folder"
      }
    }
  }
}

Using Built Executable

{
  "mcpServers": {
    "memory-decay": {
      "command": "C:\\path\\to\\PyMemoryDecayMCP\\dist\\memory-decay-mcp.exe",
      "args": [],
      "env": {
        "MEMORY_STORAGE_PATH": "C:\\path\\to\\custom\\data\\folder"
      }
    }
  }
}

Configuration

• MEMORY_STORAGE_PATH: (Optional) Environment variable to set the directory where the vector database and journal file are stored. Defaults to ./data relative to the working directory.

Features

• Store Memory: Embeds text and stores it in LanceDB.
• Recall Memory:Retrieves relevant memories based on semantic search, filtered by memory "strength" (decay function).
• Verify History: An immutable audit log (flat file) to verify facts regardless of memory decay.