██████╗ ██████╗ ██████╗ ██████╗ ██████╗ ███████╗██╗ ██╗
██╔════╝██╔══██╗██╔════╝██╔════╝██╔═══██╗██╔════╝██║ ██║
██║ ██████╔╝██║ ██║ ██║ ██║█████╗ ██║ ██║
██║ ██╔══██╗██║ ██║ ██║ ██║██╔══╝ ╚██╗ ██╔╝
╚██████╗██║ ██║╚██████╗╚██████╗╚██████╔╝███████╗ ╚████╔╝
╚═════╝╚═╝ ╚═╝ ╚═════╝ ╚═════╝ ╚═════╝ ╚══════╝ ╚═══╝
Give your AI agents persistent memory. Store, retrieve, and reason over all information—across sessions, agents, and time.
$ echo "The Problem"
LLM agents are stateless. Every conversation starts from zero.
$ cat pain_points.txt
✗ Agents forget user preferences between sessions
✗ No learning from past interactions
✗ Can't reference previous conversations
✗ Multi-agent systems can't share knowledge
✗ Context windows fill up instantly
✗ No persistent understanding of users
$ ./solve_with_cortex.sh
✓ Cortex: Memory layer initialized. Agents now remember everything.
Features
A complete memory infrastructure for AI agents—storage, retrieval, enrichment, and lifecycle management.
Combine semantic vector search with BM25 keyword matching. Get the right memory every time with automatic re-ranking.
Multiple agents share and access the same memories. Build collaborative AI systems that learn from each other.
Automatically extract people, places, organizations, and concepts. Build knowledge graphs from conversations.
AI-powered importance scoring prioritizes critical memories. Allergies and preferences always surface first.
Similar memories are automatically detected and merged. No clutter, just clean relevant context.
Replay entire conversations. Understand context and debug agent behavior with full history.
Integration
Drop-in integration with your existing AI stack. Works with OpenAI, Anthropic, LangChain, and more.
import { Cortex } from
'@cortex/sdk';
// Initialize the memory layer
const memory = new
Cortex({ apiKey: process.env.CORTEX_API_KEY });
// Store a memory
await memory.memorize({
content: 'User prefers dark mode and works in finance',
role: 'ASSISTANT', tags: ['preference', 'user-info'], });
// Recall relevant memories
const results =
await memory.recall({
query: 'what does the user prefer?', limit:
5, });
// results[0].content → "User prefers dark mode and works in finance"
How It Works
Install the SDK and initialize with your API key. Works with any LLM or agent framework.
Every message, decision, and learning is automatically stored with rich metadata.
Query memories with natural language. Hybrid search finds exactly what's relevant.
Your agent now has context from all previous interactions. It remembers everything.
Use Cases
Remember every customer interaction. No more repeating information. Personalized support from day one.
Learn user preferences over time. Schedule meetings, set reminders, and anticipate needs.
Build knowledge bases from research. Share findings across agent teams. Never lose an insight.
Remember codebase context, coding preferences, and past decisions. Better suggestions every time.
Pricing
Start free. Scale as you grow. No hidden fees.
Free
$0/mo
Forever free
For testing and small projects
Pro
$6.99/mo
or $60/year (save 28%)
For production applications
Team
$49/mo
or $490/year (save 17%)
For teams building AI together
Pro plan overages: $0.20 per 1,000 additional memories, $0.02 per 1,000 additional API calls