Cortex

Cortex - Project Memory for AI Coding Assistants

**Your AI forgets everything between sessions. Cortex gives it a brain.**

Every time you start a new AI coding session, you waste 15-30 minutes re-explaining your project. Architecture, past decisions, known bugs, conventions — all gone.

**Cortex captures everything automatically** and injects it into your next session before you type a single character.

Works with **Claude Code** | **Cursor** | **Cline** | **Copilot** | **Any MCP client**

---

Why Cortex?

Without Cortex	With Cortex
"Here's my project structure again..."	AI already knows your architecture
"We decided to use Redis because..."	Decision auto-captured with full context
"The bug was in the auth middleware..."	Bug pattern recorded, never repeated
"Don't touch that file, it's..."	Convention remembered across sessions
15-30 min context loading per session	0 min. Full context injected automatically

---

How It Works

You code with AI  -->  Cortex watches silently  -->  Memory builds automatically
                                                           |
Next session starts  <--  Context injected into CLAUDE.md  <--  Best context selected

**Install. Code. That's it.** Zero configuration needed.

---

Features

Real-Time Memory Capture

Cortex monitors your AI sessions live — not just at the end:

• Every 1 second — Watches for new messages
• Every 15 seconds — Fast local extraction (no API call)
• Every 20 messages — Deep LLM extraction in background
• On decisions/bugs detected — Immediate capture
• Status bar shows Cortex: Live during active sessions

3-Layer Memory Architecture

Inspired by how human memory works:

**Layer 1: Working Memory (hot)** — Always injected (~800 tokens)

• Last session summary, recent decisions, open problems
• Auto-injected into CLAUDE.md before every session
• Your AI reads this automatically

...