Cortex Documentation

Cortex gives any AI agent persistent memory across conversations. Save context, preferences, and decisions — then recall them semantically when needed.

Quick Start →

Get up and running in under a minute with one command.

MCP Server →

Add memory to Claude, Kiro, Cursor, or any MCP-compatible agent.

REST API →

Full HTTP API for saving, searching, and managing memories.

Python SDK →

Native Python client for programmatic memory management.

How It Works

  1. Your agent saves important context (decisions, preferences, project details) as memories.
  2. Memories are stored with semantic embeddings for intelligent retrieval.
  3. On future conversations, the agent searches for relevant memories and loads them as context.
  4. Memories are private and scoped to your API key — no data is shared between users.

Key Features

  • • Semantic search — find memories by meaning, not just keywords
  • • Automatic decay — old, unused memories fade naturally
  • • Tag organization — categorize memories for filtering
  • • Agent scoping — isolate memories per agent or share across all
  • • MCP native — works with any MCP-compatible AI client