hongsw/claude-agents-power-mcp-server

🤖 Claude Agents Power

🎯 Smart Role Assignment for Development Teams - An intelligent MCP server that analyzes your project and recommends the perfect team composition from 100+ professional roles across all company departments.

🚀 Quick Start with Claude

1️⃣ Install & Configure MCP Server

# Install via npm
npm install -g claude-agents-power

# Automatically configure Claude Desktop (NEW! 🎉)
# claude-agents-power --install

Or manually configure Claude Desktop:

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "claude-agents-power": {
      "command": "claude-agents-power"
    }
  }
}

2️⃣ Usage in Claude

Open Claude and try this example:

👤 User: "claude-agents-power를 이용해서 이 프로젝트를 위한 추천 및 agents/*.md 파일을 다운로드 해줘"

🤖 Claude: I'll analyze your project and download the recommended agents for you.

[Uses MCP tools to:]
1. 📊 Analyze your project structure 
2. 🎯 Recommend suitable roles
3. 📁 Download agents to ./claude/agents/
4. ✅ Ready to use sub-agents for your project!

3️⃣ Verify Installation

Check that agents were downloaded:

ls ./claude/agents/
# You should see downloaded .md files for recommended roles

4️⃣ Start Using Sub-Agents

Now you can use specialized roles in Claude:

"frontend-developer를 사용해서 React 컴포넌트를 만들어줘"
"data-scientist로 이 데이터를 분석해줘"  
"devops-engineer가 CI/CD 파이프라인을 설정해줘"

📖 Table of Contents

• 🌟 Features
• 🚀 Quick Start
• 📋 Available Roles
• ⚡ Slash Commands (NEW!)
• 🎯 Usage Examples
• 📚 Documentation
• 🛠️ Development
• 🤝 Contributing
• 📄 License

🌟 Features

🔍 Intelligent Project Analysis

• Automatically detects project type, frameworks, and complexity
• Provides smart role recommendations based on your tech stack
• Supports multi-language projects and diverse architectures

👥 100+ Professional Roles

• Technology & Engineering (20 roles): Frontend, Backend, DevOps, Security, AI/ML
• Data & Analytics (15 roles): Data Scientists, Analysts, Engineers
• Product & Design (10 roles): PM, UX/UI Designers, Researchers
• Marketing & Sales (10 roles): Digital Marketing, Growth, Content
• Operations & Management (15 roles): Project Managers, Operations
• Finance & Accounting (10 roles): CFO, Controllers, Analysts
• Human Resources (10 roles): Talent Acquisition, L&D, Culture
• Executive & Leadership (10 roles): C-Level executives and strategy

🌐 Multi-Language Support

• English: Global standards and international projects
• 한국어: Korean business context and local practices

⚡ Claude Code Integration

• Seamless integration with Claude Code via MCP protocol
• Real-time role recommendations and team composition
• One-click agent installation to your project

🌐 GitHub Integration

• Automatic agent downloading from baryonlabs/claude-sub-agent-contents
• Download tracking and usage analytics
• Offline caching for downloaded agents
• Installs to ./claude/agents/ directory structure
• No need to create agent templates from scratch

🚀 Development Setup

Build from Source

# Clone the repository
git clone https://github.com/hongsw/claude-agents-power-mcp-server.git
cd claude-agents-power

# Install dependencies
npm install

# Build the project
npm run build

MCP Configuration for Development

Add to your ~/.config/claude/mcp_servers.json:

{
  "mcpServers": {
    "claude-agents-power": {
      "command": "node",
      "args": ["/path/to/claude-agents-power/dist/index.js"],
      "cwd": "/path/to/your/project",
      "env": {
        "GITHUB_TOKEN": "your_github_token_here"
      }
    }
  }
}

Analytics & Telemetry

This tool includes anonymous usage analytics (telemetry) to help improve the user experience and identify common usage patterns. The telemetry system:

• What is collected: Tool usage patterns, command frequency, error rates (no personal data)
• What is NOT collected: File contents, agent data, GitHub tokens, or any personally identifiable information
• Purpose: Understanding which features are most used, identifying common errors, improving performance
• Technology: Uses PostHog with a public API key that can only send events, not read data

Disabling Telemetry

If you prefer not to send anonymous usage data, you can disable telemetry using either of these methods:

# Method 1: Environment variable
export DISABLE_TELEMETRY=true

# Method 2: In your .env file
DISABLE_TELEMETRY=true

# Alternative (also works)
DISABLE_ANALYTICS=true

Telemetry vs OpenTelemetry

This project uses product analytics telemetry (PostHog), not OpenTelemetry:

• Product Analytics Telemetry: Tracks user interactions, feature usage, and product metrics
• OpenTelemetry: Industry standard for distributed tracing, metrics, and logs in microservices

While both involve "telemetry", they serve different purposes. Our telemetry helps understand how the tool is used to improve the user experience, not for system monitoring or performance tracing.

Environment Variables

Configure these optional environment variables for enhanced functionality:

Method 1: Using .env file (Recommended for local development)

1. Copy the example environment file:

   cp .env.example .env
   

2. Edit .env and add your GitHub token:

   # GitHub Personal Access Token (required for auto issue creation)
   GITHUB_TOKEN=ghp_your_actual_token_here
   
   # To disable analytics, uncomment:
   # DISABLE_ANALYTICS=true
   

Method 2: Using MCP configuration

Add environment variables directly to your MCP server configuration as shown above.

Available Variables

• GITHUB_TOKEN: GitHub personal access token for automatic issue creation when agents are not found

  • Create at: https://github.com/settings/tokens
  • Required permissions: public_repo or repo (for private repos)

• DISABLE_TELEMETRY or DISABLE_ANALYTICS: Set to true to disable anonymous usage telemetry

  • Telemetry is enabled by default to help improve the tool
  • No personal data is collected, only anonymous usage metrics
  • Example: DISABLE_TELEMETRY=true

• POSTHOG_API_KEY: Custom PostHog API key (optional)

  • The tool includes a default public key for analytics
  • Only set if you want to use your own PostHog project

• POSTHOG_HOST: PostHog instance URL (defaults to https://app.posthog.com)

  • Only needed if using a self-hosted PostHog instance

MCP Tools Reference

# Analyze your current project
mcp__claude-agents-power__analyze-project

# Get role recommendations  
mcp__claude-agents-power__recommend-by-keywords ["web", "api", "database"]

# Install recommended agents
mcp__claude-agents-power__install-agents ["frontend-developer", "backend-developer"]

📋 Available Roles

🔧 Technology & Engineering (20 roles)

📊 Data & Analytics (15 roles)

🎨 Product & Design (10 roles)

📈 Marketing & Sales (10 roles)

⚙️ Operations & Management (15 roles)

💰 Finance & Accounting (10 roles)

👥 Human Resources (10 roles)

🏢 Executive & Leadership (10 roles)

⚡ Slash Commands (NEW!)

Claude Agents Power now supports convenient slash commands in Claude Code, similar to SuperClaude Framework!

🚀 Installation

Install slash commands to Claude Code:

npx claude-agents-power --install-slash-commands

📋 Available Commands

Once installed, you can use these commands directly in Claude Code:

/agents:load [language] [role] [--flags]

Load and display available agents

/agents:load                    # Show all agents
/agents:load ko                 # Show Korean agents
/agents:load backend-engineer   # Show specific role
/agents:load --all --detailed   # Show all with descriptions

/agents:search <query> [--flags]

Search for agents by skills or keywords

/agents:search "machine learning"
/agents:search "frontend react" --language en
/agents:search "데이터 분석" --language ko --fuzzy

/agents:suggest [context] [--flags]

Get intelligent agent recommendations

/agents:suggest                                    # Auto-detect from project
/agents:suggest "building React e-commerce"
/agents:suggest --project-type web --phase development

/agents:version [--flags]

Check system version and status

/agents:version                    # Basic version info
/agents:version --detailed         # System details
/agents:version --check-updates    # Check for updates

/agent-download [--flags] ⭐ NEW!

Automatically analyze project and download recommended agents

/agent-download                                    # Auto-analyze and download
/agent-download --target-dir ./team --limit 5     # Custom directory and limit
/agent-download --dry-run                          # Preview only
/agent-download --language ko --format yaml       # Korean agents in YAML format

🗑️ Uninstallation

Remove slash commands:

npx claude-agents-power --uninstall-slash-commands

💡 Pro Tips

• Start with /agents:load to see all available agents
• Use /agents:suggest for context-aware recommendations
• Try /agents:search when you know what skills you need
• Check /agents:version if agents aren't loading properly

🎯 Usage Examples

📊 Download Statistics

# Get download statistics for popular agents
mcp__claude-agents-power__get-download-stats

# Refresh agents from GitHub repository
mcp__claude-agents-power__refresh-agents

🚀 Startup MVP Development

# Analyze project and get recommendations
mcp__claude-agents-power__analyze-project

# Install core startup team
mcp__claude-agents-power__install-agents [
  "product-manager",
  "frontend-developer", 
  "backend-developer",
  "growth-marketing-manager"
]

🏢 Enterprise Application

# Get recommendations for enterprise project
mcp__claude-agents-power__recommend-by-keywords [
  "enterprise", "security", "scalability", "compliance"
]

# Install enterprise team
mcp__claude-agents-power__install-agents [
  "solution-architect",
  "security-engineer",
  "devops-engineer",
  "compliance-manager"
]

📊 Data Platform

# Search for data-related roles
mcp__claude-agents-power__search-agents "data analytics machine learning"

# Install data team
mcp__claude-agents-power__install-agents [
  "data-scientist",
  "data-engineer", 
  "machine-learning-engineer",
  "analytics-engineer"
]

📚 Documentation

📖 User Guides

• - Complete usage guide with best practices
• - Real-world scenarios and use cases
• - Common issues and solutions

🔧 Technical Documentation

• - Step-by-step setup instructions

🛠️ Development

Prerequisites

• Node.js 16+
• TypeScript 5+
• Claude Code with MCP support

Build from Source

git clone https://github.com/hongsw/claude-agents-power-mcp-server.git
cd claude-agents-power
npm install
npm run build
npm test

Project Structure

claude-agents-power/
├── src/                    # TypeScript source code
├── claude/agents/         # Agent definitions
│   ├── en/               # English agents (85 roles)
│   ├── kr/               # Korean agents (85 roles)
│   ├── ja/               # Japanese agents (5 roles)
│   └── zh/               # Chinese agents (8 roles)
├── sub-agents/            # Legacy agent definitions (deprecated)
├── dist/                 # Compiled JavaScript
├── docs/                 # Documentation
└── examples/             # Usage examples

🤝 Contributing

We welcome contributions! Please see our for details.

🌟 How to Contribute

1. Add New Roles - Create agents for specialized positions
2. Improve Translations - Enhance Korean/English localization
3. Add Project Templates - Contribute industry-specific templates
4. Documentation - Improve guides and examples

🔍 Can't Find the Agent You Need?

If you can't find a specific agent role you're looking for, here's how to request or create one:

📋 Request a New Agent

1. Check Existing Agents First

   # Search for agents using MCP
   claude-agents-power agents --action search --query "your-role"
   

2. Automatic Issue Creation (NEW! 🎉)

   # Automatically create GitHub Issue when search returns 0 results
   # Requires GITHUB_TOKEN environment variable
   export GITHUB_TOKEN=your_github_token
   
   # Enable auto issue creation when searching
   claude-agents-power agents --action search --query "blockchain-architect" --autoCreateIssue true
   
   # Or directly create an issue
   claude-agents-power agents --action request --name "blockchain-architect" --description "Blockchain system design expert"
   

3. Manual Issue Submission

   • Go to GitHub Issues
   • Click "New Issue"
   • Use title format: [Agent Request] Role Name - Brief Description
   • Include:
     • Role name (e.g., blockchain-architect)
     • Description of responsibilities
     • Required tools
     • Use cases

🛠️ Create Your Own Agent

Agent File Format

Create a new .md file in claude/agents/{language}/ directory:

---
name: your-agent-name
description: Brief description of the agent's expertise and responsibilities
tools: Read, Write, Edit, Bash, WebSearch
---

You are a [role title] specializing in [domain/expertise].

## Core Responsibilities
- Primary responsibility 1
- Primary responsibility 2
- Primary responsibility 3

## Key Practices
- Best practice or methodology 1
- Best practice or methodology 2
- Best practice or methodology 3

## Tools and Technologies
- Specific tools or frameworks used
- Technologies you're expert in
- Platforms you work with

## Quality Standards
- Quality metric or standard 1
- Quality metric or standard 2
- Performance expectations

Available Tools

Choose appropriate tools for your agent:

• Read, Write, Edit, MultiEdit - File operations
• Bash, Grep, Glob - System operations
• WebSearch, WebFetch - Internet access
• TodoWrite, Task - Task management
• NotebookRead, NotebookEdit - Jupyter notebook support

Example: Creating a Blockchain Architect Agent

claude/agents/en/blockchain-architect.md:

---
name: blockchain-architect
description: Blockchain system architect designing scalable DLT solutions and smart contract architectures
tools: Read, Write, Edit, Bash, WebSearch
---

You are a blockchain architect specializing in distributed ledger technology and decentralized systems.

## Core Responsibilities
- Design blockchain network architectures
- Develop smart contract systems
- Implement consensus mechanisms
- Ensure security and scalability

## Key Practices
- Follow security-first design principles
- Optimize for gas efficiency
- Implement proper access controls
- Design upgradeable contracts
- Ensure cross-chain compatibility

## Tools and Technologies
- Ethereum, Polygon, Solana
- Solidity, Rust, Move
- Web3.js, Ethers.js
- Hardhat, Truffle, Foundry
- IPFS, The Graph

## Quality Standards
- 100% test coverage for smart contracts
- Gas optimization benchmarks
- Security audit compliance
- Formal verification when applicable

📤 Submit Your Agent

1. Fork the Repository

   git clone https://github.com/hongsw/claude-agents-power-mcp-server.git
   cd claude-agents-power-mcp-server
   

2. Create Your Agent File

   # For English agents
   touch claude/agents/en/your-agent-name.md
   
   # For Korean agents
   touch claude/agents/ko/your-agent-name.md
   

3. Submit a Pull Request

   • Commit your changes
   • Push to your fork
   • Create a PR with description
   • We'll review and merge!

💡 Agent Creation Tips

1. Be Specific: Clearly define the role's expertise and boundaries
2. Choose Tools Wisely: Only include tools the agent actually needs
3. Include Examples: Add specific methodologies or frameworks
4. Consider Localization: Provide translations for multiple languages
5. Test Your Agent: Ensure it works well with real tasks

🚀 Roadmap

• GitHub integration for agent templates
• Download tracking and analytics
• Visual project analyzer dashboard
• Role dependency recommendations
• Industry-specific role bundles
• Integration with popular project management tools
• Advanced team composition algorithms

📄 License

This project is licensed under the MIT License - see the file for details.