mcp-ai-memory

This MCP server provides a semantic memory system that lets AI agents store, retrieve, and manage contextual information across sessions. It uses PostgreSQL with pgvector for vector similarity search, Redis for optional caching, and a type-safe TypeScript implementation to ensure robust memory operations. Available tools include memory_search to perform natural-language memory lookups, memory_store to persist new memories while checking for duplicates, memory_update and memory_delete for metadata and lifecycle management, and advanced operations like memory_batch for bulk imports and memory_graph_search or memory_traverse for exploring relationships in memory graphs. You can also query memory_stats for database insights and memory_consolidate to merge or cluster similar memories. Use memory_get_relations to inspect relationships between memories and memory_relate/memory_unrelate to manage connections between items.

Prerequisites:

• Node.js 18+ (or Bun) installed on your system
• PostgreSQL installed with pgvector extension enabled
• Optional Redis for caching

Install from npm (recommended):

npm install -g mcp-ai-memory

Install from source:

1. Install dependencies:

bun install

2. Set up PostgreSQL with pgvector:

CREATE DATABASE mcp_ai_memory;
\c mcp_ai_memory
CREATE EXTENSION IF NOT EXISTS vector;

3. Create environment file:

# Create .env with your database credentials
touch .env

4. Run migrations (if using source version):

bun run migrate

Usage:

• Development:

bun run dev

• Production:

bun run build
bun run start

Environment tips:

• Ensure DATABASE_URL (or MEMORY_DB_URL) points to a PostgreSQL database with the vector extension enabled.
• If Redis is unavailable, the server will fall back to in-memory caching, which may affect performance under heavy load.
• EMBEDDING_MODEL controls the embedding dimensions; mismatches between model and database can cause embedding errors. Consider pinning a model and re-embedding if changing models.
• MEMORY_DB_URL and related env vars can be customized in your deployment config or Claude Desktop integration to fit your environment.
• For large-scale deployments, monitor memory_decay_status and memory_preserve to manage lifecycle and keep important memories intact.