Skip to content
Merged
Changes from all commits
Commits
File filter

Filter by extension

Filter by extension

Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
125 changes: 110 additions & 15 deletions README.md
Original file line number Diff line number Diff line change
@@ -1,20 +1,115 @@
# Cortex — AI Orchestration System
# 🧠 Cortex

## What is Cortex?
Cortex is a personal AI corporation system that orchestrates AI agents (Jules, Codex, Claude) via GitHub Issues to complete development tasks. The human acts as CEO — setting goals, reviewing PRs, and merging, while agents perform the actual work.
[![Python 3.12](https://img.shields.io/badge/python-3.12+-blue.svg)](https://www.python.org/downloads/release/python-3120/)
[![License: Open Source](https://img.shields.io/badge/License-Open%20Source-brightgreen.svg)](https://opensource.org/)
[![PRs Welcome](https://img.shields.io/badge/PRs-welcome-brightgreen.svg)](http://makeapullrequest.com)
[![Powered by Claude Code](https://img.shields.io/badge/powered%20by-Claude%20Code-6b4fbb.svg)](https://claude.ai/)

## How it Works
Cortex follows a command-based orchestration flow to automate development:
1. **Plan**: `/council` generates tasks based on project roles (CPO, CTO, CMO, Growth).
2. **Assign**: `/dispatch` creates GitHub Issues from these tasks with appropriate agent labels.
3. **Queue**: GitHub Issues act as a task queue for the agents.
4. **Execute**: Agents like **Jules** and **Codex** pick up tasks, implement solutions, and submit Pull Requests.
5. **Review**: The human CEO reviews and merges the Pull Requests.
### Your Personal AI-Corporation
**Orchestrate a team of AI agents via GitHub Issues.**

## Usage
Cortex turns your GitHub repository into a self-evolving AI corporation. You act as the CEO—setting goals and reviewing results—while a specialized team of AI agents handles the implementation, testing, and security.

### /council
The `/council` command generates a set of tasks based on the current project context and defined roles. It analyzes the project's goals and breaks them down into actionable items for the AI agents. It also supports a **Forge Port Sprint** mode to benchmark against claude-forge and produce 3-5 concrete feature-port tasks for Cortex.
---

### /dispatch
The `/dispatch` command takes the generated tasks and creates corresponding GitHub Issues. It automatically assigns the correct labels (e.g., `jules`) so the AI agents know which tasks to pick up and work on.
## 🚀 Quick Start

Get your AI corporation up and running in less than 5 minutes:

1. **Clone the brain**:
```bash
git clone https://github.com/your-username/cortex.git
cd cortex
```
2. **Start the engine**:
Launch [Claude Code](https://claude.ai/code) in the project root.
3. **Run your first Council**:
```bash
/council
```
*The AI council (CPO, CTO, CMO) analyzes your project and suggests actionable tasks.*
4. **Dispatch tasks**:
```bash
/dispatch
```
*Cortex creates GitHub Issues and assigns them to specialized agents like **Jules** or **Codex**.*
5. **Review and Merge**:
Watch as PRs arrive. Review the code, run `/verify`, and merge!

---

## 🛠 Features

### 💻 11 Specialized Commands
Power up your CLI with custom slash commands:
- `/council`: Strategic planning and task generation.
- `/dispatch`: Automatic issue creation and agent assignment.
- `/heartbeat`: Automated tech trend scanning (HN, GitHub, Reddit).
- `/status`: Real-time health check of your agent pipeline.
- `/verify`: Automated verification of task completion.
- *And more: /handoff, /new-project, /screenshot, /tdd, /build-fix, /learn.*

### 🤖 4 AI Agents
Your dedicated workforce, ready for any task:
- **Architect**: High-level system design and tech stack decisions.
- **Code Reviewer**: Automated PR analysis and quality enforcement.
- **Security Auditor**: Scanning for vulnerabilities and secret leaks.
- **Verify Agent**: Ensuring every task meets the Definition of Done.

### ⚓ 8 Hardened Git Hooks
Built-in safety and quality checks:
- `check-secrets`: Prevent accidental leaks of API keys.
- `protect-main`: Enforce branch protection and PR workflows.
- `mcp-usage-tracker`: Monitor and optimize MCP tool usage.
- *Plus: check-filesize, pre-commit-check, grab-screenshot, output-secret-filter, expensive-tool-warning.*

### 🔄 3 Automated Workflows
Set-and-forget automation via GitHub Actions:
- **Heartbeat Cron**: Keeps your project updated with the latest AI trends.
- **Auto-Review**: Instant feedback on every Pull Request.
- **Jules Trigger**: Seamless handoff from Issue to Implementation.

---

## 🏗 Architecture

```text
[ YOU: CEO ]
|
| /council (Planning)
v
[ AI CONCILIUM ] <--- (CPO, CTO, CMO, Growth)
|
| /dispatch (Execution)
v
[ GITHUB ISSUES ] <--- (The Task Queue)
|
+------+------+
| |
[ JULES ] [ CODEX ] <--- (AI Workers)
| |
+------+------+
|
v
[ PULL REQUESTS ]
|
| (Human Review)
v
[ PRODUCTION ]
```

---

## ⚖️ Comparison

| Feature | Cortex | claude-forge | Aider / Cursor |
| :--- | :---: | :---: | :---: |
| **Orchestration** | Multi-agent via Issues | Single CLI | IDE-based |
| **Workflow** | CEO-centric (Reviewer) | Developer-centric | Developer-centric |
| **Automation** | Cron + Hooks + Actions | Hooks only | Manual |
| **Context** | Project-wide AI Roles | File-based | Session-based |

---

## 📝 License
This project is open-source. See individual file headers for specific licensing terms where applicable.