DOCSEARCHbyAlgolia
Submit Docs

Claude Code Unified Agents

repository·main·Indexed 20 days ago

https://github.com/stretchcloud/claude-code-unified-agents

A collection of 54 production-ready specialized Claude Code sub-agents for automating workflows across development, infrastructure, quality, AI, and business domains. Includes a master @orchestrator for multi-agent coordination and a variety of domain-specific agents such as @backend-architect, @devops-engineer, and @security-auditor. Features documentation on installation, agent invocation via @mentions, and custom agent configuration using Markdown and YAML frontmatter.

Tokens
7.1K
Snippets
24
Records
29
Agent score
72%

What's inside claude-code-unified-agents

  1. Implement common multi-agent workflow patterns

    main

    When designing your own agentic workflows, follow these established patterns to ensure efficiency and quality:

    • Design → Implement → Test → Deploy: A linear progression using @ux-designer → @frontend-specialist → @test-engineer → @devops-engineer.
    • Parallel Development: Executing independent tracks simultaneously, such as @backend-architect (API), @frontend-specialist (UI), and @mobile-developer (App).
    • Security-First Development: Integrating security checks at multiple stages: @security-auditor (threat model) → @backend-architect (secure implementation) → @security-auditor (final audit).
    • Data-Driven Features: Building features from the data up: @data-engineer (pipeline) → @ai-engineer (model) → @backend-architect (API) → @frontend-specialist (UI).

  2. Orchestrate complex projects using @orchestrator

    main

    To manage large-scale or multi-faceted projects, use the @orchestrator agent. The orchestrator acts as a central coordinator that receives a high-level request and decomposes it into a structured delegation plan involving specialized agents (e.g., @backend-architect, @frontend-specialist, @devops-engineer).

    Workflow Pattern:

    1. Initial Request: Provide a comprehensive goal to @orchestrator.
    2. Delegation: The orchestrator identifies necessary phases (Planning, Development, Testing, Deployment) and assigns specific agents to tasks.
    3. Execution: Agents work through the plan, often in parallel or in a chained sequence based on dependencies.
    @orchestrator Build a complete e-commerce platform with user authentication, product catalog, shopping cart, and payment integration using React and Node.js

  3. Best practices for effective multi-agent use

    main

    To get the best results from the unified agents, follow these five principles:

    1. Clear Task Definition: Be highly specific about requirements to help the orchestrator generate an optimal plan.
    2. Leverage Parallel Execution: Identify tasks that do not depend on each other so they can be executed simultaneously.
    3. Chain Dependent Tasks: Explicitly sequence tasks where one output is required for the next task's input.
    4. Use Specialized Agents: Always select the most appropriate agent for a specific domain (e.g., use @blockchain-developer for smart contracts rather than a generalist).
    5. Review and Iterate: Use specialized review agents (like @security-auditor or @test-engineer) to validate work before moving to the deployment phase.

  4. Install Claude Code Unified Agents

    main

    To install the full suite of agents, run the quick-install script via curl. This script automates the setup process for all available agents in the repository.

    # Install all agents
curl -sSL https://raw.githubusercontent.com/stretchcloud/claude-code-unified-agents/main/scripts/quick-install.sh | bash

  5. Build an MVP using the Orchestrator and specialized agents

    main

    You can use the @orchestrator to manage a multi-phase development lifecycle by delegating specific tasks to specialized agents. This pattern is effective for rapid prototyping (MVPs) where different agents handle foundation, core features, and launch phases.

    30-Day AI Writing Assistant Pattern

    Phase 1: Foundation (Days 1-10)

    • @orchestrator: Manage the project.
    • @backend-architect: Setup serverless API.
    • @prompt-engineer: Design writing prompts.
    • @frontend-specialist: Create simple UI.

    Phase 2: Core Features (Days 11-20)

    • @ai-engineer: Integrate GPT API.
    • @fullstack-engineer: User authentication.
    • @frontend-specialist: Editor interface.

    Phase 3: Launch (Days 21-30)

    • @devops-engineer: Deploy to production.
    • @monitoring-specialist: Setup analytics.
    • @content-strategist: Launch content.
    @orchestrator quick MVP for AI writing tool:
@backend-architect: Setup serverless API
@prompt-engineer: Design writing prompts
@frontend-specialist: Create simple UI

  6. Use Claude Code Unified Agents

    main

    Once installed, you can invoke the agents by using the claude command followed by a natural language instruction. The agents are designed to handle complex tasks such as building APIs, managing infrastructure, or specialized domain work.

    # Start using agents
claude "Build a REST API with authentication"

  7. Use the @orchestrator command for multi-agent projects

    main

    The @orchestrator command is used to initiate large-scale, multi-agent projects by defining high-level requirements. The orchestrator then identifies and coordinates the necessary specialized agents (e.g., fintech-specialist, ai-engineer, cloud-architect) to execute the project phases.

    Common patterns include:

    • Digital Bank Launch: Building mobile-first, AI-powered, and compliant banking systems.
    • Healthcare Ecosystems: Creating HIPAA-compliant platforms with telemedicine and AI diagnosis.
    • Metaverse/Gaming: Developing MMOs with virtual economies and blockchain integration.
    • E-Commerce: Building personalized, AR-enabled shopping experiences.
    • Autonomous Systems: Developing vehicle AI, fleet management, and telemetry processing.
    @orchestrator build a digital-only bank from scratch with:
- Mobile-first experience
- AI-powered financial advice
- Cryptocurrency support
- International transfers
- Regulatory compliance

  8. Use Claude Code Unified Agents

    main

    Once installed, you can interact with the agents in Claude Code using several methods:

    Interactive Selection

    Use the /agents command to browse and select agents manually:

    /agents

    Automatic Agent Selection

    Claude will automatically select the most appropriate agent based on your natural language request. For example, asking about REST API design will trigger the backend-architect agent.

    Explicit Agent Invocation

    Directly mention a specific agent using the @ symbol to ensure a specific expertise is used:

    @frontend-specialist create a React component for user authentication

    Multi-Agent Orchestration

    For complex, multi-step tasks, use the @orchestrator. The orchestrator analyzes requirements, creates a delegation plan, coordinates multiple agents, and integrates their results.

    @orchestrator Build a complete e-commerce platform with payment integration
    @frontend-specialist create a React component for user authentication
@orchestrator Build a complete e-commerce platform with payment integration

  9. Install Claude Code Unified Agents

    main

    You can install the agents using a quick install script or by manually copying them to your project or user directory.

    Quick Install (Recommended)

    Use the provided installation script to clone and set up the agents. You can also specify only certain agents to install.

    # Clone the repository
git clone https://github.com/stretchcloud/claude-code-unified-agents.git

# Run the installation script
cd claude-code-unified-agents
./scripts/install.sh

# Or for specific agents only
./scripts/install.sh --agents backend-architect,frontend-specialist

    Manual Installation

    Project-level agents (Local to current project)

    Create a .claude/agents directory in your project root and copy the agents there:

    mkdir -p .claude/agents
cp -r claude-code-unified-agents/claude-code-unified-agents/.claude/agents/* .claude/agents/

    User-level agents (Global for all projects)

    Create a global agents directory in your home folder:

    mkdir -p ~/.claude/agents
cp -r claude-code-unified-agents/claude-code-unified-agents/.claude/agents/* ~/.claude/agents/
    # Quick Install
git clone https://github.com/stretchcloud/claude-code-unified-agents.git
cd claude-code-unified-agents
./scripts/install.sh

# Specific agents
./scripts/install.sh --agents backend-architect,frontend-specialist

  10. Use Claude Code Unified Agents via @mentions

    main

    Agents are invoked using the @ symbol followed by their unique identifier.

    Simple Agent Call

    To perform a task with a single specialized agent, use the format: @agent-name <task>.

    Example:

    @python-pro optimize this function for performance

    Complex Task with Orchestrator

    The orchestrator agent acts as a master coordinator for complex multi-agent workflows. Use it to manage high-level goals that require multiple specialized agents.

    Example:

    @orchestrator build a real-time chat application with video calling

  11. Execute Enterprise Digital Transformation projects

    main

    For large-scale enterprise transformations, the @orchestrator can manage a multi-phase roadmap involving different agent groups at each stage:

    1. Phase 1: Assessment: Uses business-analyst, requirements-analyst, and cloud-architect to deliver a current state analysis and roadmap.
    2. Phase 2: Foundation: Uses cloud-architect, devops-engineer, and security-auditor to build cloud infrastructure and security frameworks.
    3. Phase 3: Modernization: Uses backend-architect, database-specialist, and data-engineer to migrate to microservices and data platforms.
    4. Phase 4: Innovation: Uses ai-engineer, frontend-specialist, and mobile-developer to implement AI capabilities and modern interfaces.
    @orchestrator lead digital transformation for Fortune 500 company:
- Legacy system modernization
- Cloud migration
- Process automation
- Data analytics platform
- Employee training

  12. Implement industry-specific solutions via @orchestrator

    main

    You can target specific industries by providing specialized requirements to the @orchestrator. The system will pull in relevant agents for that domain.

    Financial Services

    • Robo-advisor: Requires fintech-specialist, ai-engineer, data-scientist, frontend-specialist, and security-auditor.
    • Crypto Exchange: Requires blockchain-developer, fintech-specialist, security-auditor, and performance-engineer.
    • Insurtech: Requires ai-engineer, fintech-specialist, mobile-developer, and data-scientist.

    Healthcare

    • Clinical Trials: Requires healthcare-dev, data-engineer, ai-engineer, and security-auditor.
    • Mental Health: Requires healthcare-dev, ai-engineer, mobile-developer, and ux-designer.
    • Medical Imaging AI: Requires ai-engineer, healthcare-dev, data-engineer, and security-auditor.

    Retail & E-Commerce

    • Omnichannel Retail: Requires ecommerce-expert, backend-architect, mobile-developer, and data-engineer.
    • Social Commerce: Requires frontend-specialist, backend-architect, ai-engineer, and fintech-specialist.
    • B2B Marketplace: Requires ecommerce-expert, backend-architect, fintech-specialist, and workflow-optimizer.

    Gaming & Entertainment

    • Cloud Gaming: Requires game-developer, cloud-architect, performance-engineer, and frontend-specialist.
    • Esports: Requires game-developer, backend-architect, ai-engineer, and fintech-specialist.
    • Music Streaming: Requires backend-architect, ai-engineer, mobile-developer, and data-engineer.

    Manufacturing & Logistics

    • Supply Chain: Requires backend-architect, ai-engineer, data-engineer, and embedded-engineer.
    • Factory Automation: Requires embedded-engineer, ai-engineer, data-engineer, and monitoring-specialist.
    • Last-Mile Delivery: Requires backend-architect, ai-engineer, mobile-developer, and monitoring-specialist.
    @orchestrator build robo-advisor platform:
- Portfolio optimization
- Risk assessment
- Tax-loss harvesting
- Goal planning
- Rebalancing