G0t4 MCP Server Memory File

🚀 MCP Server Memory

This is an MCP server that interacts with a memory text file to assist Claude in cross-chat context management.

Each line represents a piece of memory.

These tools enable Claude (and other MCP clients) to manage memories during a chat:

• memory_add(memory: string) - Add a memory.
• memory_search(query: string) - Return matching memories (exact substring match) - Wildcards/regular expressions may be supported in the future.
• memory_delete(query: string) - Delete matching memories (exact substring match).
• memory_list() - Return all memories.
• Note: memory_update == memory_delete + memory_add.

For example:

• I mention my name => "Chatting with Wes".
• Mention my daughter's age => "Wes's daughter is 8 years old".
• Say I'm working on a TypeScript project => "Working on a TypeScript project".
• Moreover, it's important to base on what Claude (the assistant/LLM) says or does...
  • Especially tool usage (i.e., run_command). For instance, when the first attempt to use a tool fails (e.g., the python command doesn't exist), and then python3 is used successfully later => Claude can record "Used python3, python was not available".
• I ask Claude to delete memories about X => memory_delete(query: X).
• I correct my name => memory_search("oldname") + Delete each matching item or common subset query + Add the new name.

Then, when starting a new chat, Claude will automatically retrieve the most recent memories (subset or all) or can request memories (partial/more/all). These memories can then influence responses, tool usage, etc.

✨ Features

📚 Design

A simple memory text file is used for the following reasons:

• ChatGPT's memory works well and is essentially a text file.
  • Although it may be structured behind the scenes, when you view the memories, they are presented as a text file.
• It worked well when I tested it in the mcp-server-commands for a similar reminder feature (when Claude had access to them).
• Unstructured text simplifies the management of tools and parameters, essentially managing a list of strings.

💡 Prompt Mechanism

• It's crucial to have a prompt to let the model know when to store memories. However, it's not clear how to do this best...
• Training: OpenAI acknowledges that some training teaches the model when to store memories, similar to how the model is trained for tool usage.
• Prompt: The system prompt section may include a reminder to trigger memory storage.
• The tools themselves: In my tests with Claude, I couldn't get it to store memories based solely on tool specifications or prompts/suggestions in tool responses. So, relying solely on tools is not enough. I believe adding a prompt in the prompt section would be effective.

🔮 Future Plans/Ideas

I'm not sure if these are worth the effort. I'm just listing some future ideas, maybe to stop myself from implementing them :)

• Freshness factor: A way to reorder memories based on freshness?
  • Order becomes relevant in this case, especially for fuzzy memory queries (e.g., when working on a TypeScript project and a Python project, and I ask to start a new project, it might suggest the one used most recently).
• Gradually fade out old memories?