Memory3
Verified IntegrationClient Configuration
— Connect Memory3 to Claude Desktop or Cursor in seconds{
"mcpServers": {
"memory3": {
"command": "npx",
"args": [
"-y",
"@modelcontextprotocol/server-memory3"
],
"env": {}
}
}
}~/Library/Application Support/Claude/claude_desktop_config.json (macOS) or %APPDATA%\Claude\claude_desktop_config.json (Windows).System Overview
Provides an ultra-lightweight, fully local AI memory system designed for Model Context Protocol (MCP) clients, featuring semantic storage and hybrid search capabilities.
7/23/2026
Open Source
stdio / SSE RPC
Frequently Asked Questions
Architecture and operational details for Memory3
Memory3 offers 'Hybrid Search (RRF)', which integrates vector semantics with BM25 keyword matching for superior search precision. For speed, it uses 'HNSW Indexing' to enable ultra-fast approximate nearest neighbor searches, ensuring low-latency retrieval even for large memory sets.
Related MCP Servers
Browse all servers →Empower AI assistants to generate and convert documents, manage templates, and automate document workflows efficiently.
Provides a multi-tenant, AI-native Content Delivery Network deployable on Cloudflare, featuring sub-100ms TTFB, AI agent controllability, and comprehensive accessibility.
Deploy a Model Context Protocol server on Cloudflare Workers without requiring authentication.