Memory Custom

What is Memory Custom?

Memory Custom is an enhanced version of the official MCP Memory server. It adds custom memory paths and timestamping capabilities to the knowledge graph that captures interactions via a language model (LLM). It's intended for users who want project-specific memory storage and better context tracking.

How to use Memory Custom?

Install via Smithery with npx -y @smithery/cli install @BRO3886/mcp-memory-custom --client claude or clone the repo, run npm install && npm run build, and set the MEMORY_FILE_PATH environment variable to specify a custom memory file path (defaults to memory.json). Add the server configuration to claude_desktop_config.json or .cursor/mcp.json and follow the provided system prompt for interaction.

Key features of Memory Custom

• Custom memory paths for project-specific storage
• Timestamping for interaction tracking and context
• Create, read, update, and delete entities and relations
• Add observations with timestamps
• Search and open nodes in the knowledge graph
• Read entire graph structure

Use cases of Memory Custom

• Manage separate memory files for different projects
• Track history of interactions with precise timestamps
• Organize knowledge graph by project contexts
• Enhance LLM context with structured, timestamped memory

FAQ from Memory Custom

What are the prerequisites for running Memory Custom?

Node.js version 16 or higher is required.

How do I set a custom memory file path?

Set the MEMORY_FILE_PATH environment variable before running the server. If not set, it defaults to memory.json in the same directory as the script.

What tools does Memory Custom expose?

It exposes tools for setting memory file path, creating/reading/updating/deleting entities, relations, observations, searching nodes, opening nodes, and getting the current time.

How does timestamping work?

The server provides a get_current_time tool that can be used to add timestamps to entities, observations, and relations during interaction.

Can I use Memory Custom with Claude Desktop or Cursor?

Yes, it can be configured as an MCP server in claude_desktop_config.json or .cursor/mcp.json by specifying the command and args to the compiled index.js file.