mcp-server-memory
by MCP-Mirror
mcp-server-memory is an MCP server designed to interact with a memory text file, enabling Claude to maintain inter-chat context. It allows Claude and other MCP clients to manage memories mid-chat by adding, searching, deleting, and listing them.
Last updated: N/A
What is mcp-server-memory?
This is an MCP server that interfaces with a simple text file to store and manage memories for Claude or other MCP clients. It provides tools to add, search, delete, and list memories, allowing the LLM to retain context across multiple chat sessions.
How to use mcp-server-memory?
The server exposes several functions: memory_add(memory: string) to append a memory, memory_search(query: string) to find matching memories, memory_delete(query: string) to remove memories, and memory_list() to retrieve all memories. Claude can use these tools to store important information from conversations and retrieve them later to influence responses or tool usage.
Key features of mcp-server-memory
Memory addition
Memory search (substring exact match)
Memory deletion (substring exact match)
Memory listing
Simple text file storage
Use cases of mcp-server-memory
Maintaining user preferences across chat sessions
Remembering details about ongoing projects
Storing information about tool usage successes and failures
Correcting and updating information
Providing context for new chats
FAQ from mcp-server-memory
How does the server store memories?
How does the server store memories?
The server stores memories as individual lines within a text file.
What is the cueing mechanism for storing memories?
What is the cueing mechanism for storing memories?
The server relies on a combination of training, prompt engineering (system prompt component), and potentially tool use to cue the model when to store memories.
How does the search function work?
How does the search function work?
The search function performs a substring exact match to find relevant memories.
How can I update a memory?
How can I update a memory?
Updating a memory involves deleting the old memory using memory_delete and then adding the new memory using memory_add.
What are some potential future improvements?
What are some potential future improvements?
Potential future improvements include adding a recency factor to prioritize recent memories and implementing a mechanism to fade out old memories.