A Model Context Protocol (MCP) server that enables Large Language Models to store and retrieve persistent memories with intelligent search capabilities.
MemFlow MCP provides seamless integration between LLMs like Claude and your Memory Bank API, allowing for persistent memory management across conversations. The server supports adding memories with tags, semantic search, and flexible memory retrieval.
npx memflow-mcpOr install globally:
pnpm add -g memflow-mcpMEMBANK_API_URL=http://localhost:3000
MEMBANK_API_KEY=your-api-key # optionalSee CLAUDE_DESKTOP_CONFIG.md for detailed configuration options for different Node.js installations.
Quick start - Find your npx path and use it:
which npxThen configure Claude Desktop:
{
"mcpServers": {
"memflow": {
"command": "/your/npx/path",
"args": ["-y", "memflow-mcp"],
"env": {
"MEMBANK_API_URL": "http://localhost:3000"
}
}
}
}- addMemory - Store content with optional tags
- searchMemory - Search memories with semantic matching
- listMemories - Browse stored memories with filtering
Once configured, you can use these commands in Claude:
Add this to memory: "Claude can now remember things across conversations"
Search my memories for "conversations"
List my recent memories
- Node.js 18+
- Memory Bank API server running
-
Find your npx path:
which npx
Use this full path in your Claude Desktop config.
-
Common paths:
- asdf:
/Users/username/.asdf/shims/npx - Homebrew (Intel):
/usr/local/bin/npx - Homebrew (Apple Silicon):
/opt/homebrew/bin/npx - System:
/usr/bin/npx
- asdf:
-
Test it works:
/your/npx/path -y memflow-mcp
- Ensure
MEMBANK_API_URLis set correctly - Check that your Memory Bank API is running
- Restart Claude Desktop after config changes
Your Memory Bank API should support:
POST /memory- Create memoryGET /memory- List/search memories
MIT