MCP Memory Custom

🚀 Custom Memory Server

This project enhances the memory server provided by the MCP team with new features. It allows users to create and manage knowledge graphs for capturing interactions with Large Language Models (LLMs).

✨ Features

1. Custom Memory Path

• Users can now specify different memory file paths for different projects.
• Why? This feature enhances the ability to organize and manage memory data, supporting project-specific storage.

2. Timestamp Functionality

• The server now generates timestamps for interactions.
• Why? Timestamps enable users to track when each memory was created or modified, providing better context for the data's history.

🚀 Quick Start

Prerequisites

• Node.js (version 16 or higher)

Installation via Smithery

Automatically install the knowledge graph memory server for Claude Desktop using Smithery:

npx -y @smithery/cli install @BRO3886/mcp-memory-custom --client claude

Manual Installation

1. Clone the repository:

   git clone git@github.com:BRO3886/mcp-memory-custom.git
   cd mcp-memory-custom
   

2. Install dependencies:

   npm install
   

Configuration

Before running the server, you can set the MEMORY_FILE_PATH environment variable to specify the path of the memory file. If not set, by default, the server will use memory.json in the same directory as the script.

Running the Server

Update the MCP Server JSON File

Add the following to your claude_desktop_config.json or .cursor/mcp.json file:

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

System prompt changes:

Follow these steps for each interaction:
1. The memory file path for this project is /path/to/memory/project_name.json - Always pass this path in all memory file operations involving file manipulation (e.g., when creating entities, relationships, or retrieving memories).
2. Display this content before the user input and wait for the user's confirmation.
3. Provide this content along with the user's input to the language model.
4. After receiving the language model's response, present the response content in a user-readable manner.
5. Ensure these steps are always followed in all interactions.

Example

User input:

Hello!

System prompt displays content and waits for confirmation:

You entered "Hello!". Please confirm the following information is accurate:
- Memory file path: /path/to/memory/project_name.json
- Current time: [Current time]
- User identity: Claude 2

After user confirmation, provide this content along with the user's input to the language model.

📚 Documentation

API Endpoints

Tool List

Here are some examples of available tools:

1. Get the current time: /api/time
2. Set the memory file path: /api/memory/path
3. Create an entity: /api/entity/create
4. Delete an entity: /api/entity/delete
5. Update a relationship: /api/relationship/update

System Prompt

The system prompt is the guiding rule for user interactions with the language model. Here is an example:

Follow these steps for each interaction:
1. User input: Get the current time.
2. The system displays content and waits for confirmation.
3. Pass this content to the language model.
4. After the language model returns a response, present the result to the user.
5. Always follow these steps in all interactions.

Example

User input:

What's the current time?

System prompt displays content and waits for confirmation:

You entered "What's the current time?". Please confirm the following information is accurate:
- Memory file path: /path/to/memory/project_name.json
- Current time: [Current time]

After user confirmation, pass this content to the language model.

📄 License

This project is inspired by the memory server functionality of Anthropic.