Back to Mind & Machine
BMAD Method
Your AI Team for the Full Development Lifecycle
What is BMAD?
BMAD (Breakthrough Method of Agile AI Driven Development) is like having a complete software team in your IDE. Instead of asking AI to "write code," you collaborate with specialized agents who guide you through the entire development lifecycle.
Traditional AI: "Write me a login function"
BMAD: "Let's analyze requirements, design the auth flow, consider security implications, and implement it together"
BMAD: "Let's analyze requirements, design the auth flow, consider security implications, and implement it together"
100% free and open source. No paywalls. No gated content. Just AI-powered development that actually works.
BMAD in the SDLC
BMAD covers every phase of software development. Each phase has dedicated agents and workflows:
Complete Development Lifecycle
Planning
Product Brief
Analysis
PRD
Design
Architecture
Implementation
Development
Testing
TEA Module
Deployment
DevOps
How It Works in Practice
1
Planning Phase
PM Agent helps you write a Product Brief. You discuss the problem, target users, and success metrics. Output: A 1-page brief everyone understands.
2
Analysis Phase
PM + UX Agents create detailed requirements. User stories, acceptance criteria, edge cases. Output: A complete PRD with user flows.
3
Design Phase
Architect Agent designs the technical solution. Database schema, API contracts, service boundaries. Output: An Architecture Decision Record.
4
Implementation & Testing
Developer + QA Agents write code and tests together. TDD, code review, continuous validation. Output: Production-ready code with tests.
Meet Your AI Team
BMAD gives you 12+ specialized agents, each with their own expertise and personality. They're not just tools—they're collaborators:
Alex (PM)
Requirements & Strategy
Maya (Architect)
System Design
Jordan (UX)
User Experience
Sam (Developer)
Implementation
Taylor (QA)
Testing & Quality
Casey (Scrum Master)
Process & Agile
Plus: DevOps, Security, Data, Storyteller, Game Designer, and more!
Choose Your Path
BMAD adapts to your needs. Pick the approach that fits your task:
Quick Flow
For: Bug fixes, refactoring, small features
quick-spec→ Plan in minutesquick-dev→ Implement immediately- No ceremony, just results
- Auto-escalation if scope grows
Full Method
For: New products, major features, architecture
- Product Brief → Requirements
- PRD → Architecture Document
- Epic/Story Breakdown
- Complete traceability
Smart Escalation: Start with Quick Flow. If BMAD detects the work is bigger than expected, it suggests switching to Full Method. Your work carries forward automatically.
Party Mode 🎉
Sometimes you need the whole team in one room. Run party-mode and get all your agents collaborating:
Post-Mortem: Authentication Outage
You
Our auth service went down 3 times this sprint. What happened?
Maya (Architect)
The design had proper fallbacks. If implementation deviated from the spec, that's where we should look...
Sam (Developer)
I followed the architecture doc exactly. But it didn't account for concurrent session invalidation race conditions.
Alex (PM)
Both of you missed the bigger issue—we didn't validate session management requirements in the PRD. This is on me for not catching it during analysis.
Taylor (QA)
And I should've caught this in integration tests. We need to add concurrent invalidation scenarios to our test suite.
Perfect for: Architecture decisions, brainstorming, post-mortems, sprint planning—anytime you need diverse perspectives.
Specialized Modules
Extend BMAD with modules for specific domains:
BMAD Method (BMM)
Core framework with 34+ workflows
BMAD Builder (BMB)
Create custom agents & workflows
Test Architect (TEA)
Risk-based testing & automation
Game Dev Studio (BMGD)
Unity, Unreal, Godot workflows
Creative Intelligence (CIS)
Innovation & design thinking
Get Started in 30 Seconds
$ npx bmad-method install
Follow the prompts, then open your AI IDE. That's it! You're ready to BMAD.
Lost? Just ask:
/bmad-help tells you exactly what's next. Try /bmad-help I finished architecture, what now?How the Workflow Works
BMAD orchestrates the entire development process through a structured workflow. Here's what happens under the hood:
1. Task Intake & Classification
When you describe what you want to build, BMAD automatically:
- Analyzes scope — Is this a bug fix, feature, or new product?
- Identifies stakeholders — Who needs to be involved? PM, Architect, QA?
- Chooses the path — Quick Flow or Full Method
2. Agent Handoff
Each phase triggers specialized agents:
1
Planning
PM Agent captures requirements, user stories, success metrics
2
Analysis
UX + PM Agents create flows, edge cases, acceptance criteria
3
Design
Architect Agent designs schema, APIs, service boundaries
4
Build
Developer Agent writes code; QA Agent writes tests in parallel
3. Continuous Validation
At every step, BMAD validates against previous artifacts:
- Code → Architecture — Does implementation match design?
- Tests → Requirements — Do tests cover acceptance criteria?
- Changes → Impact — What else might break?
4. Artifact Generation
BMAD automatically creates and maintains:
- Product Brief — 1-page problem statement
- PRD — Detailed requirements with user flows
- Architecture Decision Records — Technical design docs
- Code — Implementation with tests
- Changelog — What changed and why
The key difference: BMAD doesn't just generate code—it generates the entire development conversation, keeping every decision traceable from idea to deployment.
Ready to Try BMAD?
Get started in seconds:
$ npx bmad-method install
Follow the prompts, then open your AI IDE. You're ready to BMAD.
Story: Implementing a New Feature
Let's follow a story: implementing push notifications for a SaaS app. See BMAD in action:
Round 1: The Idea
You
I want to add push notifications to our app. Users should get notified when someone mentions them in a comment.
Alex (PM Agent)
Great idea! Before we dive in, let me clarify a few things:
- Who sends notifications? (other users, system, or both)
- Do users need to opt-in/out?
- What channels? (email, push, in-app, or all?)
You
Other users mention each other with @. System sends for billing alerts. Yes for opt-in. All three channels.
Alex (PM Agent)
Got it! This is a Full Method feature—multiple stakeholders involved. Let me create the Product Brief.
Output: Product Brief
docs/briefs/2025-02-21-push-notifications.md
docs/briefs/2025-02-21-push-notifications.md
Round 2: Requirements Analysis
Alex (PM Agent)
Based on our chat, here are the user stories:
- As a user, I want to mention @others so they get notified
- As a user, I want to enable/disable notifications per channel
- As a user, I want to see notification history
- As a system, I want to send billing alerts
Jordan (UX Agent)
I've designed the notification preferences screen. Users can toggle: Email, Push (browser), In-App. Each has a preview of what it looks like.
You
Looks good! What about email formatting?
Jordan (UX Agent)
Added email templates. Light HTML with inline styles for maximum email client compatibility.
Output: PRD (Product Requirements Document)
docs/prd/2025-02-21-push-notifications.md — 12 user stories, acceptance criteria, edge cases
docs/prd/2025-02-21-push-notifications.md — 12 user stories, acceptance criteria, edge cases
Round 3: Architecture Design
Maya (Architect Agent)
Based on the PRD, I'm designing the notification system:
- Notification Service — handles queuing, preferences
- Event Bus — listens for @mentions, billing events
- Channel Handlers — email, push, in-app各有职责
- User Preferences Table — stores opt-in settings
You
What about scaling? We might send thousands at once.
Maya (Architect Agent)
Great question! I'm adding:
- Async queue (Redis + Bull)
- Rate limiting per user (max 10/min)
- Retry with exponential backoff
- Dead letter queue for failed sends
Alex (PM Agent)
Adding to traceability: I've linked these technical decisions to specific user stories. Every line of code will be accountable.
Output: Architecture Decision Records
docs/adr/2025-02-21-notification-service.md — API contracts, DB schema, queue design
docs/adr/2025-02-21-notification-service.md — API contracts, DB schema, queue design
Round 4: Implementation
Sam (Developer Agent)
Starting implementation. I've broken this into 4 PRs:
- Database migrations + UserPreferences model
- Notification Service core + queue setup
- @mention event handler
- Channel handlers (email, push, in-app)
You
How's the first PR looking?
Sam (Developer Agent)
Done! Created:
20250221_add_user_preferences.sqlmodels/notification_preferences.py- API endpoints for preferences CRUD
Maya (Architect Agent)
I've verified: implementation matches the ADR. The queue design uses Bull as specified. Good to merge.
DB Migration
Service
Event Handler
Channels
Round 5: Testing
Taylor (QA Agent)
I've created the test plan based on the PRD acceptance criteria:
- Unit tests: 45 scenarios
- Integration tests: 12 user journeys
- Edge cases: @mention in edit, deleted user mentions
Taylor (QA Agent)
Found one issue: when a user is mentioned in a deleted comment, the system crashes. Adding edge case handler now.
You
Good catch! What about performance under load?
Taylor (QA Agent)
Load tested: 1000 mentions/min with 99.7% success rate. The rate limiter kicks in correctly at 10/min per user.
All tests passing! 157 unit tests, 23 integration tests, 0 critical bugs
Delivered!
You
Ready to ship!
Alex (PM Agent)
Here's the delivery summary:
- 4 PRs merged
- All user stories complete
- Documentation updated
- Monitoring alerts configured
Alex (PM Agent)
Added to CHANGELOG.md:
## [2.3.0] - 2025-02-21
### Added
- Push notifications for @mentions
- Email and in-app notification channels
- User notification preferences panel
- Billing alert notifications
### Changed
- Comment system to support @mentions
### Added
- Push notifications for @mentions
- Email and in-app notification channels
- User notification preferences panel
- Billing alert notifications
### Changed
- Comment system to support @mentions
Feature shipped in 3 days!
Every decision traceable. Every test passing. Every stakeholder aligned.