Enhanced BMAD Method with Context Management for Claude Code

This repository contains the Enhanced BMAD Method - combining persona-based AI agents with advanced context management, persistent memory, and comprehensive testing orchestration for Claude Code.

🚀 What's New

The Enhanced BMAD Method integrates:

• Persistent Context Management - Agents maintain memory across sessions
• Comprehensive Testing Orchestration - Zero-regression development
• Agent Health Monitoring - Track performance and optimize workflows
• Dynamic Agent Creation - System creates specialized agents as needed
• Cross-Project Learning - Knowledge preserved and shared

📁 Project Structure

.claude/
├── agents/              # Enhanced BMAD agent definitions
├── context/             # Persistent agent memory
├── test-specs/          # Testing specifications
├── workflows/           # Development workflows
├── agent-analytics/     # Performance metrics
├── templates/           # Reusable patterns
└── commands/            # Utility scripts

🤖 Available Agents

Core Orchestration

• bmad-orchestrator - Master orchestrator with context management, health monitoring, and dynamic agent creation

Planning Phase Agents (Personas)

• bmad-analyst (Mary) - Business Analyst for market research and project briefs
• bmad-pm (John) - Product Manager for PRDs and user stories
• bmad-architect (Winston) - System Architect for technical design
• bmad-po (Sarah) - Product Owner for validation and alignment

Development Phase Agents

• bmad-sm (Bob) - Scrum Master for detailed story creation
• bmad-dev (James) - Full Stack Developer for implementation
• bmad-qa (Quinn) - QA Architect for code review

Supporting Agents

• bmad-ux (Sally) - UX Designer for interfaces and user experience
• bmad-devops (Alex) - DevOps Engineer for infrastructure
• bmad-tech-writer (Morgan) - Technical Writer for documentation

Infrastructure Agents

• testing-orchestrator - Comprehensive testing coordination
• project-initializer - Set up existing projects with BMAD

🚀 Quick Start

New Projects

# Start with workflow guidance
Use bmad-orchestrator for workflow guidance

# Or begin with research
Use bmad-analyst to research and create project brief

Existing Projects

# Initialize BMAD + Context system
Use project-initializer to set up this codebase

📋 Enhanced Workflow

Phase 1: Planning (Web UI)

1. Research → bmad-analyst creates project brief
2. Product → bmad-pm creates PRD and stories
3. Architecture → bmad-architect designs system
4. Validation → bmad-po ensures alignment

Phase 2: Development (IDE)

1. Stories → bmad-sm creates detailed stories
2. Code → bmad-dev implements features
3. Review → bmad-qa ensures quality
4. Testing → testing-orchestrator validates everything

Phase 3: Deployment

1. Infrastructure → bmad-devops sets up CI/CD
2. Documentation → bmad-tech-writer creates docs

🎮 Orchestrator Commands

The enhanced bmad-orchestrator supports:

• *help - Show available agents and workflows
• *agent [name] - Transform into specific agent
• *workflow-guidance - Get workflow recommendations
• *context-status - View agent contexts and health
• *agent-health - Display performance metrics
• *test-orchestrate - Start comprehensive testing
• *init-project - Initialize project with BMAD
• *create-agent [type] [name] - Create specialized agent

🧠 Context Management

Persistent Memory

Each agent maintains context files:

• Research findings (Mary)
• Product decisions (John)
• Architecture patterns (Winston)
• Code patterns (James)
• Review feedback (Quinn)

Agent Health Tracking

# .claude/agent-analytics/health-metrics.yaml
agent_health:
  mary-analyst:
    usage_frequency: 12/week
    success_rate: 95%
    health_status: "healthy"

🧪 Testing Integration

Comprehensive Coverage

• Unit Tests
• Integration Tests
• Workflow Tests (end-to-end)
• Performance Tests
• Accessibility Tests (WCAG)
• Console Error Monitoring

Quality Gates

• Zero console errors policy
• Performance budgets enforced
• Accessibility compliance required
• Test coverage thresholds

📊 Advanced Features

Dynamic Agent Creation

The orchestrator monitors patterns and creates specialized agents when needed:

• Connector agents for complex integrations
• Domain specialists for new technologies
• Performance optimizers for bottlenecks

Cross-Project Learning

• Successful patterns saved as templates
• Knowledge preserved when agents retire
• Best practices shared across projects

Tmux Integration

For 24/7 autonomous operation:

• .claude/commands/start-bmad-project.sh
• .claude/commands/schedule-check-in.sh
• .claude/commands/send-agent-message.sh

🛠️ Installation

1. Claude Code automatically detects agents in .claude/agents/
2. Each agent is a Markdown file with YAML frontmatter:

---
name: agent-name
description: Agent purpose and capabilities
tools: Read, Write, Edit, Task, TodoWrite
---

Agent system prompt...

🏃 Usage Examples

Complete Feature Development

"Add user authentication to my app"
# Orchestrator coordinates full workflow:
# Analyst → PM → Architect → Dev → QA → Testing

Quick Implementation

"Implement user profile API"
# Goes directly to Dev (James)

Quality Check

"Review and test the auth system"
# QA (Quinn) + Testing Orchestrator

📚 Best Practices

1. Trust the Process - Let agents handle their expertise
2. Review Context - Check .claude/context/ for insights
3. Monitor Health - Use *agent-health regularly
4. Test Everything - Comprehensive testing prevents regressions
5. Learn and Adapt - System improves with usage

🔧 Customization

Modify Agents

1. Edit files in .claude/agents/
2. Adjust prompts for your team
3. Update tool access in frontmatter
4. Add custom commands

Create New Agents

1. Use *create-agent command
2. Or manually create in .claude/agents/
3. Follow BMAD persona conventions
4. Initialize context file

🚨 Troubleshooting

• Agent Not Found: Restart Claude Code to reload
• Commands Not Working: Use * prefix
• Context Issues: Check .claude/context/
• Test Failures: Review .claude/test-reports/

🔗 Resources

• Original BMAD Method
• Claude Code Sub-Agents
• Context Management Patterns

📝 Key Differences from Original BMAD

1. Persistent Memory - Agents remember across sessions
2. Health Monitoring - Track and optimize performance
3. Dynamic Creation - System creates agents as needed
4. Testing Integration - Comprehensive quality assurance
5. Cross-Project Learning - Knowledge preservation

License

Based on the BMAD Method (MIT License)