@technioz/claude-agents

Multi-platform, multi-agent orchestration system - Complete SDLC automation with 9 specialized agents for Claude Code, Cursor, and more

🎉 What's New in v1.2.0

Multi-Platform Support! 🚀

• ✅ Claude Code & Cursor - Now supports both platforms with a single codebase
• ✅ Platform Registry - Scalable architecture for easy addition of future platforms (Codex, etc.)
• ✅ Platform Flags - Use --claude, --cursor, or --platform with all commands
• ✅ Coexistence - Install agents for multiple platforms in the same project
• ✅ Backward Compatible - Defaults to Claude Code, no breaking changes

View Changelog | Full Release Notes

🚀 What is Claude Agents?

Claude Agents is a powerful multi-platform, multi-agent orchestration system designed to work with Claude Code, Cursor, and other AI-powered development environments. It provides a complete Software Development Life Cycle (SDLC) automation through 9 specialized AI agents that work together to build, test, secure, and deploy your applications.

Supported Platforms

• 🤖 Claude Code - Anthropic Claude integration (default)
• 🖱️ Cursor - Cursor IDE integration
• More platforms coming soon...

The 9 Agents

1. ARCHITECT - Design system architecture, component structure, and data flow
2. DEVELOPER - Write clean, type-safe code implementing designs
3. DESIGNER - Create UI components using shadcn/ui
4. INTEGRATIONS - Manage external API connections (OpenAI, Anthropic, Groq, etc.)
5. TESTING - Ensure code quality through automated testing
6. SECURITY-and-PRODUCTION-READINESS - Security review and production readiness validation
7. DEPLOYMENT - Manage production deployment and CI/CD
8. DOCUMENTATION - Keep documentation current and comprehensive
9. ORCHESTRATOR - Master controller that delegates tasks to all agents

📦 Installation

Quick Start

# Install globally
npm install -g @technioz/claude-agents

# Or use npx (no installation required)
npx @technioz/claude-agents init

Initialize Agents in Your Project

# Interactive setup (recommended) - will prompt for platform
claude-agents init

# Install for Claude Code (default)
claude-agents init --claude

# Install for Cursor
claude-agents init --cursor

# Install all agents locally
claude-agents init --local

# Install all agents globally
claude-agents init --global

# Install specific agents
claude-agents init --agents ARCHITECT DEVELOPER DESIGNER

# Install for Cursor with specific agents
claude-agents init --cursor --agents ARCHITECT DEVELOPER

# Skip prompts and install all agents (defaults to Claude Code)
claude-agents init --yes

🎯 Features

• ✅ Multi-Platform Support - Works with Claude Code, Cursor, and more
• ✅ Complete SDLC Coverage - From architecture to deployment
• ✅ Interactive CLI - User-friendly setup wizard
• ✅ Flexible Installation - Global or project-specific agents
• ✅ Selective Installation - Choose which agents you need
• ✅ Easy Updates - Keep agents up to date with one command
• ✅ Custom Agents - Create your own specialized agents
• ✅ Validation - Ensure agent files are correctly formatted
• ✅ Comprehensive Documentation - Includes AGENTS_PROTOCOL.md

📚 Usage

Available Commands

init - Initialize Agents

Install agents in your project or globally.

# Interactive setup (prompts for platform)
claude-agents init

# Platform selection
claude-agents init --claude          # Use Claude Code (default)
claude-agents init --cursor          # Use Cursor
claude-agents init --platform cursor # Specify platform explicitly

# Installation location
claude-agents init --global          # Install globally
claude-agents init --local            # Install locally

# Agent selection
claude-agents init --agents A B C    # Install specific agents
claude-agents init --yes             # Skip prompts, install all

# Combined example
claude-agents init --cursor --local --agents ARCHITECT DEVELOPER

list - List Agents

View all available or installed agents.

# List all available agents
claude-agents list

# List installed agents (shows all platforms if no platform specified)
claude-agents list --installed

# List installed agents for specific platform
claude-agents list --installed --claude
claude-agents list --installed --cursor

update - Update Agents

Update existing agents to the latest version.

# Interactive update (prompts for platform)
claude-agents update

# Update for specific platform
claude-agents update --claude
claude-agents update --cursor

# Update global agents
claude-agents update --global

# Update local agents
claude-agents update --local

# Update specific agents
claude-agents update --agents ARCHITECT DEVELOPER

# Combined example
claude-agents update --cursor --local --agents ARCHITECT

create - Create Custom Agent

Create a new custom agent from template.

# Interactive creation (prompts for platform)
claude-agents create MY-CUSTOM-AGENT

# Create for specific platform
claude-agents create MY-CUSTOM-AGENT --claude
claude-agents create MY-CUSTOM-AGENT --cursor

# Create in global directory
claude-agents create MY-CUSTOM-AGENT --global

# Create in local directory
claude-agents create MY-CUSTOM-AGENT --local

# Combined example
claude-agents create MY-CUSTOM-AGENT --cursor --local

validate - Validate Configuration

Validate agent configuration files.

# Validate local agents (defaults to Claude Code)
claude-agents validate

# Validate for specific platform
claude-agents validate --claude
claude-agents validate --cursor

# Validate global agents
claude-agents validate --global

# Validate local agents
claude-agents validate --local

# Validate specific path (overrides platform)
claude-agents validate --path /path/to/agents

🏗️ Agent Architecture

Development Pipeline Flow

1. ARCHITECT → Design system architecture
2. DESIGNER → Create UI component specifications
3. DEVELOPER → Implement components
4. INTEGRATIONS → Build API providers (can run parallel with #3)
5. TESTING → Write and run tests
6. SECURITY-and-PRODUCTION-READINESS → Security review (MANDATORY GATE)
   ├─ APPROVED → Proceed to #7
   ├─ CONDITIONAL → Fix issues, re-review
   └─ REJECTED → Fix critical issues, re-review
7. DOCUMENTATION → Document features
8. DEPLOYMENT → Deploy to production (ONLY after security approval)

Communication Protocol

All agents communicate using a structured JSON protocol:

• TASK_ASSIGNMENT - ORCHESTRATOR delegates work
• TASK_ACKNOWLEDGMENT - Agent confirms receipt
• TASK_COMPLETION - Agent reports finished work
• TASK_REJECTION - ORCHESTRATOR requests fixes
• BLOCKER_ESCALATION - Agent reports blockers
• BLOCKER_RESOLUTION - ORCHESTRATOR resolves blockers

See AGENTS_PROTOCOL.md for complete protocol specification.

📖 Documentation

File Structure

After installation, agents are stored in platform-specific directories:

Claude Code (default):

.claude/
├── agents/
│   ├── ARCHITECT.md
│   ├── DEVELOPER.md
│   ├── DESIGNER.md
│   ├── INTEGRATIONS.md
│   ├── TESTING.md
│   ├── SECURITY-and-PRODUCTION-READINESS.md
│   ├── DEPLOYMENT.md
│   ├── DOCUMENTATION.md
│   └── ORCHESTRATOR.md
└── AGENTS_PROTOCOL.md

Cursor:

.cursor/
├── agents/
│   ├── ARCHITECT.md
│   ├── DEVELOPER.md
│   └── ... (same agent files)
└── AGENTS_PROTOCOL.md

You can have agents installed for multiple platforms simultaneously. Each platform maintains its own directory structure.

Agent Metadata

Each agent includes:

• name - Agent identifier
• description - Purpose and role
• model - AI model to use (sonnet, opus, haiku)
• color - Display color in Claude Code

Customization

You can customize agents by editing their .md files:

1. Modify the agent's Purpose and Duty
2. Update Instructions for specific behavior
3. Add or remove Capabilities
4. Define Limits and constraints
5. Customize Communication Protocol responses

🎨 Creating Custom Agents

Create specialized agents for your specific needs:

claude-agents create DATA-ANALYST

The CLI will guide you through:

1. Setting description
2. Choosing AI model
3. Selecting color
4. Configuring capabilities

Then edit the generated file to add:

• Specific instructions
• Domain knowledge
• Integration patterns
• Success criteria

🔒 Security Gate

The SECURITY-and-PRODUCTION-READINESS agent acts as a mandatory quality gate:

What it checks:

• ✅ Security vulnerabilities (SQL injection, XSS, CSRF)
• ✅ Production readiness (env vars, error handling, logging)
• ✅ Scalability (database optimization, caching, rate limiting)
• ✅ Code robustness (error handling, retry logic, timeouts)
• ✅ Dependency vulnerabilities

Approval Status:

• APPROVED - Code can be deployed
• CONDITIONAL - Minor issues, fix before production
• REJECTED - Critical issues, deployment blocked

Critical Rule: NO deployment without SECURITY_AUDITOR approval!

💡 Examples

Example 1: Full Feature Development

# Initialize all agents for Claude Code
claude-agents init --yes

# Or for Cursor
claude-agents init --cursor --yes

# Start your IDE (Claude Code or Cursor) and request a feature
# "Build a chat interface with model selector"

# Agents will automatically:
# 1. ARCHITECT designs the system
# 2. DESIGNER creates UI specs
# 3. DEVELOPER implements code
# 4. INTEGRATIONS adds API providers
# 5. TESTING writes tests
# 6. SECURITY reviews everything
# 7. DOCUMENTATION documents it
# 8. DEPLOYMENT deploys to production

Example 2: Security Review Only

# Install just security agent for Claude Code
claude-agents init --agents SECURITY-and-PRODUCTION-READINESS

# Or for Cursor
claude-agents init --cursor --agents SECURITY-and-PRODUCTION-READINESS

# Use in your IDE for security audits

Example 3: Update Workflow

# Check for new agent versions
claude-agents list --installed

# Update all agents for Claude Code
claude-agents update --claude --yes

# Update all agents for Cursor
claude-agents update --cursor --yes

# Validate after update
claude-agents validate --claude
claude-agents validate --cursor

Example 4: Multi-Platform Setup

# Install for both platforms in the same project
claude-agents init --claude --local --agents ARCHITECT DEVELOPER
claude-agents init --cursor --local --agents ARCHITECT DEVELOPER

# List agents from all platforms
claude-agents list --installed

# Work with platform-specific commands
claude-agents update --claude --local
claude-agents update --cursor --local

🛠️ Development

Local Development

# Clone repository
git clone https://github.com/technioz/claude-agents.git
cd claude-agents

# Install dependencies
npm install

# Link for local testing
npm link

# Test CLI
claude-agents --help

Running Tests

npm test

Linting

npm run lint

📄 License

MIT License - see LICENSE file for details

🔗 Links

• GitHub: https://github.com/technioz/claude-agents
• npm: https://www.npmjs.com/package/@technioz/claude-agents
• Issues: https://github.com/technioz/claude-agents/issues
• Claude Code: https://claude.com/claude-code
• Cursor: https://cursor.sh

🙏 Acknowledgments

• Built for use with Claude Code and Cursor
• Inspired by multi-agent systems and SDLC automation
• Uses shadcn/ui design patterns

📊 Project Stats

• 🤖 9 Specialized Agents
• 🌐 Multi-Platform Support (Claude Code, Cursor, and more)
• 📦 Zero Configuration needed
• 🔄 Complete SDLC coverage
• ✅ Production Ready
• 🏗️ Scalable Architecture - Easy to add new platforms

🚀 Roadmap

• [x] Multi-platform support (Claude Code & Cursor) - ✅ v1.2.0
• [ ] Additional platform support (Codex, etc.)
• [ ] Agent marketplace for community templates
• [ ] Visual workflow designer
• [ ] Integration with popular CI/CD platforms
• [ ] Real-time collaboration features
• [ ] Performance analytics dashboard
• [ ] Multi-language support

💬 Support

• 📧 Email: [email protected]
• 💬 GitHub Discussions: https://github.com/technioz/claude-agents/discussions
• 🐛 Bug Reports: https://github.com/technioz/claude-agents/issues

Made by developers for developers

Automate your SDLC with intelligent AI agents!