🚀 Claude Code Orchestra: Your Elite AI Development Squad

March 18, 2026 · View on GitHub

License: MIT Agents Teams

Build faster. Ship better. Code smarter.
This project provides a collection of 47 specialized AI agents, turning a capable generalist like Claude into a world-class engineering organization. Stop trying to teach one AI everything and instead leverage a team of pre-built experts.

🎯 The Core Idea: A Team of Specialists, Not One Generalist

While a single large language model is powerful, complex projects demand deep, nuanced expertise. This repository gives you that specialized team out-of-the-box.

  • Backend architects who deeply understand microservices.
  • Security experts who find vulnerabilities you didn't know you had.
  • Performance engineers who optimize every millisecond.
  • Framework specialists who are masters of Django, Rails, React, and Vue.

This collection of agents is the primary focus. You can use them with your own orchestration logic or use our provided example to get started quickly.

👥 Your AI Development Team

This repository contains 47 Specialized Agents organized into 10 professional teams. Explore the teams below to see the capabilities at your disposal.

Team Directory

🧭 Orchestration Team

Your project managers and workflow coordinators.

🏗️ Architecture Team

System designers who build solid foundations.

💻 Development Team

Core builders with framework expertise.

🛡️ Quality Assurance Team

Guardians of code quality and security.

🚀 DevOps & Infrastructure Team

Deployment and operations experts.

✍️ Language Experts Team

Deep language-specific mastery.

🎨 Design Team

User experience and interface experts.

📝 Content & Documentation Team

Clear communication and documentation.

🔬 Specialized Domains Team

Industry and domain experts.

🛠️ Specialized Tools Team

Platform and tool specialists.

🚀 Quick Start

1. Clone the Repository

git clone https://github.com/0ldh/claude-code-agents-orchestra.git
cd claude-code-agents-orchestra

2. Use Agents Directly (The Simple Way)

You can invoke any agent directly to leverage their specialized skills for focused tasks. This is the simplest way to use the agent collection without any setup.

# Ask the TypeScript expert to refactor a function for better type safety
claude "@typescript-expert Please refactor this function to be more type-safe: [paste your code here]"

# Get the security auditor to review a specific file
claude "@security-auditor Review this file for potential security vulnerabilities: src/auth/session.js"

3. Explore Your Agent Team

Familiarize yourself with the agents listed above. Click on any agent to see their detailed prompt and capabilities.

4. (Optional) Use the Example Orchestration Framework

This repository includes an advanced orchestration framework in CLAUDE.md.example. See the Orchestration Example section for details on how to use it.

🤔 Two Ways to Use This Agent Team: Direct Calls vs. Orchestration Framework

You can use this agent collection in two ways. Choose the approach that best fits your task.

AspectDirect Agent CallsOrchestration Framework Usage
When to UseFor simple, focused, single-step tasks. (e.g., refactoring a function, reviewing a file for security).For complex, multi-step missions requiring collaboration between multiple agents. (e.g., building a new feature, migrating a system).
Who OrchestratesYou are the orchestrator. You manually select and call the appropriate agent.Claude becomes the orchestrator, guided by the rules in CLAUDE.md.example. It automatically plans and delegates tasks.
ProsNo setup required. Fast and direct for simple jobs.Automates complex planning and coordination, significantly reducing your workload.
ConsYou are responsible for managing all steps and context for complex tasks.Requires a one-time setup of copying CLAUDE.md.example to ~/.claude/CLAUDE.md.

📂 Project Structure

All agents are organized into professional teams within the agents/ directory.

/
│
├── agents/
│   ├── 🧭 orchestration/      # Project Leadership & Planning
│   ├── 🏗️ architecture/       # System Foundation Design
│   ├── 💻 development/        # Core Feature Implementation
│   ├── 🛡️ quality-assurance/  # Quality & Security Audits
│   ├── 🚀 devops-infra/       # Deployment & Infrastructure
│   ├── ✍️ language-experts/    # Language-Specific Expertise
│   ├── 🎨 design/             # User Experience & Interface
│   ├── 📝 content-docs/       # Technical Documentation
│   ├── 🔬 specialized-domains/ # Niche Industry Expertise
│   └── 🛠️ specialized-tools/    # Specific Platform & Tool Mastery
│
├── CLAUDE.md.example         # Advanced: Example AI Rule Book
├── LICENSE
└── README.md

Agent Orchestra Index: Passive Context for Retrieval-Led Selection

The CLAUDE.md.example includes an Agent Orchestra Index inspired by Vercel's passive-context guidance: keep a small, always-available index in the root context, then retrieve the full specialist prompt only when it is relevant.

Why This Exists

The goal is not to inline every agent prompt into CLAUDE.md.example.

The goal is to keep a compressed directory of:

  1. Concrete retrieval paths to the full prompt files under ./agents/...
  2. Lightweight routing hints that help identify candidate specialists
  3. A retrieval-led workflow: find the right file, read it, then apply it

This keeps the passive context compact while still making specialist prompts cheap to find and hard to overlook.

Maintenance Workflow

The index in CLAUDE.md.example is maintained as a static example block.

When you add, rename, move, or materially rewrite agent prompts, update that index in the same change so the listed paths and routing hints stay aligned with the current catalog.

👑 Orchestration Example: Using CLAUDE.md.example

⚠️ Important: The following sections and the CLAUDE.md.example file demonstrate one possible way to orchestrate this team of specialist agents. This is a powerful but optional framework. You are encouraged to adapt it, build your own, or integrate these agents into your existing workflow.

The provided framework turns Claude into a project manager that can automatically triage tasks, delegate to the correct specialists, and manage complex workflows, all while waiting for your approval.

How to Apply the Example Framework

# First, backup your existing global rules, if any
cp ~/.claude/CLAUDE.md ~/.claude/CLAUDE.md.backup

# Apply the orchestration framework
cp CLAUDE.md.example ~/.claude/CLAUDE.md

Now, Claude will use the rules in the example file to manage the agent team in this project.

Example in Action

With the framework active, interactions look like this:

Simple Task:

You: "Fix the login bug"

Claude: "📋 EXECUTION PLAN
         ════════════════
         Classification: Simple
         Assigned to: @debugger
         Plan: Analyze login flow, identify bug, implement fix
         ⚠️ AWAITING APPROVAL"

You: "approved"
Claude: "🔄 @debugger investigating login issue..."

Complex Mission:

You: "Build a real-time chat system"

Claude: "📋 EXECUTION PLAN
         ════════════════
         Classification: Complex Mission
         Assigned to: @tech-lead-orchestrator
         ...
         Agent List:
         @api-architect, @backend-architect, @react-expert, @security-auditor
         ⚠️ AWAITING APPROVAL"

You: "proceed"
Claude: "Executing approved blueprint... 🔄 Step 1/4..."

Framework Details & Real-World Examples

Click to expand for detailed framework logic and examples

The framework is designed around a clear command structure and execution flow to ensure efficient and safe operation.

🎯 Model Optimization Strategy

The framework intelligently assigns models based on cognitive demands:

  • Opus Model (Strategic Roles): Used for complex system design, multi-agent coordination, security analysis, and architectural planning. (13 agents)
  • Sonnet Model (Implementation Roles): Used for feature development, framework-specific code, tool integration, and content creation. (34 agents)

This optimization delivers faster execution for implementation tasks while maintaining deep analytical capability for strategic decisions.

🔄 How It Works: The Enhanced Execution Framework

Phase 1: Intelligent Triage & Authorization When you submit a request, Claude (acting as gatekeeper):

  1. Classifies complexity - Simple task (1-2 files) or complex mission (3+ files).
  2. Delegates to specialists - Routes to the appropriate agent or the tech-lead-orchestrator.
  3. Presents execution plan - Shows you exactly what will happen.
  4. Waits for approval - No execution without your explicit "approved" or "proceed".

Phase 2: Coordinated Execution with Progress Tracking

  • Critical Protocol: All agent communication goes through the Task tool, the Tech Lead only creates blueprints (never executes), and Claude manages all execution after your approval.

Phase 3: Enhanced Quality Assurance

  • Built-in quality gates include model-optimized reviews, parallel execution for independent tasks, and automatic error recovery.

Phase 4: Smart Agent Authority

  • A clear command structure ensures specialists stay in their lane and the tech-lead-orchestrator focuses only on planning.

🎬 Real-World Examples

Example 1: E-Commerce Payment System

  • You: "Add Stripe checkout to my online store with subscription support"
  • Result: A secure, PCI-compliant payment system with subscription management, built by a coordinated team of Code Archaeologist, API Architect, Security Auditor, Payment Integration expert, Backend Architect, Django Expert, React Expert, and Test Automator.

Example 2: Mobile App Performance Crisis

  • You: "Our React Native app is freezing on the product list screen"
  • Result: App performance improved by 400% after Debugger, Mobile Developer, Performance Engineer, React Expert, and Database Optimizer diagnose and fix an infinite re-render loop and N+1 query problem.

Example 3: Legacy System Modernization

  • You: "Migrate our 10-year-old PHP monolith to microservices"
  • Result: A successful migration to 12 microservices with 99.9% uptime, orchestrated by the Tech Lead and executed by Legacy Modernizer, Code Archaeologist, Cloud Architect, Backend Architect, API Architect, DevOps Engineer, Laravel Expert, Database Optimizer, Test Automator, and Security Auditor.

🔧 Troubleshooting

Problem: Agents aren't coordinating or the tech-lead is trying to execute code directly. Solution: Ensure you have applied the orchestration framework from CLAUDE.md.example to your global ~/.claude/CLAUDE.md file.

Problem: Conflicts with existing CLAUDE.md instructions. Solution: Carefully merge the key "ORCHESTRATION", "PROTOCOL", and "EXECUTION" sections from our example into your existing rules.

📄 License

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