Cortex

Memory infrastructure for AI agents that actually remembers.

Cortex gives your AI agents persistent memory with automatic extraction of beliefs, learnings, entities, and commitments. It's like giving your AI a brain that learns and grows with every conversation.

What makes Cortex different?

Most memory systems just store and retrieve text. Cortex goes further:

• Cognitive Memory - Automatically extracts beliefs, learnings, and preferences from conversations
• Knowledge Graph - Builds entity relationships and tracks how they evolve over time
• Proactive Intelligence - Generates nudges, relationship health scores, and commitment tracking
• Time Travel - Query what was true at any point in time, not just now

Quick Example

import { CortexClient } from '@cortex/memory';
 
const cortex = new CortexClient({ apiKey: 'YOUR_API_KEY' });
 
// Store a memory
await cortex.memories.create({
  content: "User mentioned they're switching from React to Vue for their next project",
  source: "chat"
});
 
// Cortex automatically extracts:
// - Belief: "User prefers Vue over React for new projects"
// - Entity: "React" (technology, mentioned negatively)
// - Entity: "Vue" (technology, mentioned positively)
// - Learning: "User is exploring new frontend frameworks"
 
// Later, recall context
const context = await cortex.recall({
  query: "What frontend framework should I recommend?",
  includeProfile: true
});
// Returns: User prefers Vue, has React experience, exploring new options

Get Started

Core Concepts

• Memories - Raw information stored with automatic processing
• Beliefs - Stable facts about the user extracted from memories
• Learnings - Insights and patterns discovered over time
• Entities - People, places, things with tracked relationships
• Commitments - Promises and obligations with due dates

SDKs

• TypeScript - npm install @cortex/memory
• Python - pip install cortex-memory
• MCP Server - Works with Claude Desktop, Cursor, and other MCP clients