Memory Management

The Memory MCP server enables assistants to remember information and recall it later.

How it Works

The memory system uses AI-powered storage with mem0 AI and a Qdrant vector database for intelligent information retrieval.

Main Functions

Storage

• add_memories: Store new information
  • Automatic processing and categorization
  • Context-based storage
  • Supports structured and unstructured data

Search

• search_memories: Search stored information
  • Semantic search (finds similar concepts)
  • Full-text search
  • Relevance-based results

Retrieval

• get_all_memories: Display all memories
  • Chronological overview
  • Time period filtering possible

Update

• update_memory: Edit existing memories
  • Corrections and additions
  • Version control of changes

Storage Modes

Personal Mode (PERSONAL)

• Memories are user-specific
• Each user has their own memory
• Ideal for personal assistants

Team Mode (TEAM)

• Memories are stored per assistant
• All users of an assistant share the memory
• Perfect for team assistants and shared knowledge

Usage Examples

Personal Preferences

"Remember that I eat vegetarian"
"What was my preferred programming language again?"

Project Information

"Store the API endpoints for the new project"
"What were the requirements from the last meeting?"

Customer Data

"The customer prefers communication via email"
"Which products has this customer already purchased?"

Benefits

• Context-aware memory across sessions
• Intelligent linking of related information
• Privacy through separated storage areas
• No manual categorization needed