Cortex AI

AI collaboration brain for coding assistants - Memory, structure, and intelligence for your AI.

English | 繁體中文 | Documentation

---

The Problem

AI coding assistants are powerful, but they share critical limitations:

Limitation	Impact
No Memory	Every session starts fresh. Past decisions are forgotten.
No Structure	Responses vary wildly. Same prompt, different quality.
No Validation	No quality gates. AI can generate inconsistent code.
No Impact Awareness	AI doesn't understand how changes affect your codebase.

Result: Unpredictable "vibe coding" that requires constant human oversight.

The Solution

Cortex AI extends any MCP-compatible AI assistant with memory, structure, and intelligence.

┌─────────────────────────────────────────────────────────────────┐
│                    Your AI Assistant                            │
│              (Claude Code / Cursor / Copilot)                   │
└─────────────────────────┬───────────────────────────────────────┘
                          │ MCP Protocol
                          ▼
┌─────────────────────────────────────────────────────────────────┐
│                      CORTEX AI                                  │
│  ┌──────────────┐ ┌──────────────┐ ┌──────────────┐            │
│  │   Memory     │ │  Workflows   │ │   Analysis   │            │
│  │   System     │ │   System     │ │   Engine     │            │
│  │              │ │              │ │              │            │
│  │ • Patterns   │ │ • Spec       │ │ • Security   │            │
│  │ • Decisions  │ │ • Plan       │ │ • Quality    │            │
│  │ • Lessons    │ │ • Tasks      │ │ • Impact     │            │
│  └──────────────┘ └──────────────┘ └──────────────┘            │
└─────────────────────────────────────────────────────────────────┘

---

Features

Persistent Memory

Your AI finally remembers. Patterns, decisions, and lessons persist across sessions.

# AI learns from this session
learn "Use React Query for server state, Zustand for client state"

# Months later, AI recalls this automatically
context "state management"
# → Returns: "Previously decided: React Query for server state..."

Structured Workflows

Transform chaos into predictable delivery with phase-based development.

spec → clarify → plan → review → tasks → implement → release

Impact Analysis

AI understands your codebase structure before making changes.

impact-analyze files:["src/auth/service.ts"]
# → 23 files affected
# → 2 breaking changes detected
# → Risk level: HIGH

Code Intelligence (72 Tools)

Security scanning, quality analysis, test coverage, architecture validation—all built-in.

security-scan path:"src/"           # Find vulnerabilities
quality-analyze path:"src/"         # Detect code smells
test-suggest path:"src/services/"   # Suggest missing tests

Constitution-Driven Development

Define your project's principles once. Every AI response respects them.

<!-- .cortex/constitution.md -->
## Principles
- All functions must have JSDoc comments
- No console.log in production code
- Tests required for all public APIs

---

Quick Start

Install

npm install -g @rikaidev/cortex

Connect to Your AI Assistant

Claude Code:

claude mcp add cortex -- npx -y @rikaidev/cortex@latest start

Cursor:

// .cursor/mcp.json
{
  "mcpServers": {
    "cortex": {
      "command": "npx",
      "args": ["-y", "@rikaidev/cortex@latest", "start"]
    }
  }
}

Initialize Your Project

cd your-project
cortex init

Start Building

# In your AI assistant
spec "Add user authentication with OAuth"

---

Tool Categories

Cortex provides 72 specialized tools across 16 categories:

Category	Tools	Purpose
Workflow	spec clarify plan review tasks implement	Structured development
Memory	learn context correct	Persistent knowledge
Impact	impact-analyze impact-preview impact-validate	Change assessment
Quality	quality-analyze quality-complexity quality-duplicates	Code smell detection
Security	security-scan security-detect-secrets security-check-deps	Vulnerability scanning
Testing	test-analyze test-quality test-suggest	Test gap identification
Architecture	arch-validate arch-check arch-suggest	Pattern validation
Documentation	doc-analyze doc-validate doc-missing	Documentation quality

---

Example Session

# 1. Define what you're building
> spec "Add real-time notifications with WebSocket"
✓ Created workflow: 001-realtime-notifications
✓ Generated: spec.md, checklists/requirements.md

# 2. Clarify ambiguities
> clarify
? Should notifications persist when user is offline?
> Yes, queue and deliver on reconnect
✓ Updated: spec.md

# 3. Create technical plan
> plan
✓ Generated: plan.md, CONTEXT.md
✓ Memory updated with tech stack decisions

# 4. Break into tasks
> tasks
✓ Generated: tasks.md
✓ Identified 3 parallelizable task groups

# 5. Implement
> implement
✓ Pre-checks: .gitignore validated, dependencies checked
✓ Executing tasks with quality validation...

# 6. Release
> release
✓ Analyzed: 12 commits, 3 workflows
✓ Generated: CHANGELOG.md, RELEASE_NOTES.md

---

Before & After

Before Cortex AI

Developer: "Add authentication"
AI: *generates code*
Developer: "Wait, that's inconsistent with our patterns"
AI: *generates different code*
Developer: "You broke the user service"
AI: "I don't see how they're connected"

After Cortex AI

Developer: "Add authentication"

Cortex:
├─ Memory: "Auth pattern uses JWT + refresh tokens"
├─ Impact: "Will affect: UserService, 12 routes"
├─ Constitution: "Must include rate limiting"
└─ Security: "Checking for auth vulnerabilities..."

AI: *generates consistent, validated code*

---

Project Structure

.cortex/
├── constitution.md          # Your project's principles
├── memory/
│   └── index.json          # Persistent knowledge store
├── templates/              # Workflow templates
│   └── commands/           # AI execution guides
└── workflows/
    └── 001-feature/        # Individual workflows
        ├── spec.md
        ├── plan.md
        └── tasks.md

---

Requirements

• Node.js 18+
• MCP-compatible AI assistant (Claude Code, Cursor, or any MCP client)

---

Contributing

Contributions are welcome! See CONTRIBUTING.md.

---

License

MIT License - see LICENSE.

---

About RikaiDev

Rikai (理解) means "understanding" in Japanese. We build tools that help developers understand their work through AI assistance.

• Cortex AI - AI collaboration brain for coding assistants
• Toki - Automatic time tracking for developers

---

Stop vibe coding. Start building with intelligence.