memory

Overview

The Memory MCP Server provides a robust implementation of persistent memory using a local knowledge graph architecture. It allows AI models like Claude to remember information about users across multiple conversations by storing entities, relations, and observations. The server exposes a comprehensive set of tools for creating and managing entities, relations, and observations within the knowledge graph, enabling sophisticated memory management. The implementation is well-structured with clear separation between entities (nodes), relations (edges), and observations (node attributes), providing flexibility in representing complex knowledge structures.

Try asking AI

After installing, here are 5 things you can ask your AI assistant:

• create_entities

  Create multiple new entities in the knowledge graph

  Try:Create a new person entity named John SmithTry:Add a company entity called Acme Corp
• create_relations

  Create multiple new relations between entities

  Try:Create a relation between John Smith and Acme CorpTry:Connect Alice to Microsoft as an employee
• add_observations

  Add new observations to existing entities

  Try:Add that John speaks Spanish to his profileTry:Note that Acme Corp is a tech company
• delete_entities

  Remove entities and their relations

  Try:Remove the temporary event entityTry:Delete the outdated project record
• delete_observations

  Remove specific observations from entities

  Try:Remove that John prefers morning meetingsTry:Delete outdated observation about the company
• delete_relations

  Remove specific relations from the graph

  Try:Remove John Smith's relation to Acme CorpTry:Delete the outdated partnership between companies
• read_graph

  Read the entire knowledge graph

  Try:Show me all the information in memoryTry:Display the complete knowledge graph
• search_nodes

  Search for nodes based on query

  Try:Find all entities related to AITry:Search for anyone who works at Anthropic
• open_nodes

  Retrieve specific nodes by name

  Try:Show me information about John SmithTry:Retrieve the Acme Corp entity

Comparable tools

Installation

{
  "mcpServers": {
    "memory": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-memory"
      ]
    }
  }
}

{
  "servers": {
    "memory": {
      "command": "npx",
      "args": [
        "-y",
        "@modelcontextprotocol/server-memory"
      ]
    }
  }
}

{
  "mcpServers": {
    "memory": {
      "command": "docker",
      "args": ["run", "-i", "-v", "claude-memory:/app/dist", "--rm", "mcp/memory"]
    }
  }
}

FAQ

What storage format does the memory server use?

The server uses JSONL (JSON Lines) format for storing the knowledge graph data by default. The storage path can be customized using the MEMORY_FILE_PATH environment variable.

How does the server handle memory persistence across different AI model sessions?

The memory server maintains a persistent knowledge graph that survives between sessions. Entities, relations, and observations are stored in a local file and loaded when the server starts, allowing AI models to access accumulated information from previous interactions.

On Hacker News

Recent discussion from the developer community.

• Modelcontextprotocol/Servers: Model Context Protocol Servers▲ 1
  Story by vinnyglennon · 2025-03-20

Compare memory with