home / mcp / optimized memory mcp server
Provides a persistent knowledge-graph memory store accessible via MCP clients for Claude interactions.
Configuration
View docs{
"mcpServers": {
"agentwong-optimized-memory-mcp-server": {
"command": "docker",
"args": [
"run",
"-i",
"--rm",
"mcp/memory"
]
}
}
}This MCP server provides a persistent memory layer based on a local knowledge graph to help Claude remember user-specific information across chats. It includes tools to create entities, relations, and observations, along with practical examples and a deployment path that can run locally via Docker or through NPX/JavaScript tooling. This enables a developer-friendly workflow for testing memory-based interactions in Claude projects.
Use this MCP server with a compatible MCP client to store and retrieve persistent memory. The server exposes a knowledge graph of entities, relations, and observations that can be queried, updated, and expanded as conversations progress. Typical usage patterns include creating entities for recurring people or organizations, linking them with relations, and attaching atomic observations that describe facts about those entities. Clients can read the full graph or target specific nodes to retrieve context for a more personalized interaction.
With Claude Desktop, configure an MCP memory server in claude_desktop_config.json to choose between local execution (stdio) methods. You can run memory locally via Docker or via NPX, depending on your preferred workflow.
Prerequisites: ensure you have Docker installed if you plan to run the memory server via Docker, or Node.js with NPX if you prefer the NPX approach.
The README provides two explicit MCP configurations you can use directly in your Claude Desktop setup. Use either the Docker-based stdio configuration or the NPX-based stdio configuration.
{
"mcpServers": {
"memory": {
"command": "docker",
"args": ["run", "-i", "--rm", "mcp/memory"]
}
}
}{
"mcpServers": {
"memory": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-memory"
]
}
}
}If you plan to build the MCP server image locally, the README shows a Docker build command you can run to produce the memory image before deployment.
docker build -t mcp/memory -f src/memory/Dockerfile .The memory server implements a knowledge graph with entities, relations, and observations. It is designed to be persistent and easily queried by MCP clients. When used with Claude, it can help personalize interactions by remembering user identities, preferences, goals, and relationships across sessions.