Universal AI agent framework for agentic agile-driven development with progressive disclosure documentation optimized for AI agents
npx @tessl/cli install tessl/npm-bmad-method@4.44.1Universal AI agent framework for agentic agile-driven development. Transforms traditional software projects through specialized AI agent teams with a two-phase approach: Agentic Planning (web platforms) and Context-Engineered Development (IDEs).
interface PackageInfo {
name: 'bmad-method';
type: 'npm';
language: 'JavaScript';
install: 'npx bmad-method install';
version: '4.44.3';
nodeVersion: '>=20.10.0';
}# Install BMad Method
npx bmad-method install # Interactive mode with prompts
npx bmad-method install --full # Full installation
npx bmad-method install --ide cursor # Configure specific IDE
# Check status
npx bmad-method status # Show installation status
# List available resources
npx bmad-method list:expansions # List expansion packs
npx bmad-method list:agents # List all agentsAfter installation, agents are available in .bmad-core/ directory. Load agents in your IDE or web platform, activate with *help command.
BMad Method eliminates planning inconsistency and context loss through two distinct phases:
Phase 1: Agentic Planning (Web UI)
Phase 2: Context-Engineered Development (IDE)
Agent System: 10 specialized AI agent personas (Orchestrator, Analyst, PM, Architect, UX Expert, PO, SM, Dev, QA)
Task System: 23 tasks (21 core tasks + 2 common shared tasks) for reusable executable procedures
Workflow System: 6 structured workflows (greenfield/brownfield variants for fullstack, service, UI)
Template System: 13 YAML-based document templates with sophisticated markup language
Agent Teams: Pre-configured bundles for different project types (fullstack, ide-minimal, no-ui, all)
Expansion Packs: Domain-specific extensions (game development, creative writing, DevOps)
| Agent ID | Name | Phase | Key Commands | Use Case |
|---|---|---|---|---|
bmad-orchestrator | BMad Orchestrator | All | *agent, *workflow, *plan | Master coordinator, workflow guidance |
analyst | Mary | Planning | *brainstorm, *create-project-brief | Market research, requirements |
pm | - | Planning | *create-prd, *update-prd | Product requirements, PRD |
architect | - | Planning | *create-architecture | System design, architecture |
ux-expert | - | Planning | *create-frontend-spec, *generate-ui-prompt | UI/UX specifications |
po | Sarah | Planning/Dev | *shard-doc, *validate-story-draft | Document sharding, validation |
sm | Bob | Development | *draft | Story creation with full context |
dev | James | Development | *develop-story, *run-tests | Code implementation |
qa | Quinn | Development | *review, *gate | Quality review, gates |
Detailed Documentation: Agent Reference
| Workflow ID | Type | Agents | Use Case |
|---|---|---|---|
greenfield-fullstack | New | 8 agents | Full-stack app from scratch |
greenfield-service | New | 7 agents (no UX) | Backend/API only |
greenfield-ui | New | Frontend-focused | Frontend apps only |
brownfield-fullstack | Existing | Skip Analyst | Extend full-stack apps |
brownfield-service | Existing | Skip Analyst, no UX | Extend backend services |
brownfield-ui | Existing | Skip Analyst | Extend frontend apps |
Detailed Documentation: Workflow Reference
| Category | Key Tasks | Primary Users |
|---|---|---|
| Document Management | shard-doc, create-doc, index-docs | PO, PM, Architect |
| Story/Epic | create-next-story, create-brownfield-story, review-story | SM, PM, PO |
| Development | apply-qa-fixes, correct-course | Dev, SM, PO |
| Quality Assurance | qa-gate, test-design, trace-requirements | QA |
| Research | create-deep-research-prompt, advanced-elicitation | Analyst |
| Integration | generate-ai-frontend-prompt, kb-mode-interaction | UX, Orchestrator |
Detailed Documentation: Task Reference
| Template ID | Output | Used By | Purpose |
|---|---|---|---|
project-brief-tmpl | docs/project-brief.md | Analyst | Initial overview |
prd-tmpl | docs/prd.md | PM | Greenfield requirements |
brownfield-prd-tmpl | docs/brownfield-prd.md | PM | Enhancement requirements |
architecture-tmpl | docs/architecture.md | Architect | Generic architecture |
fullstack-architecture-tmpl | docs/fullstack-architecture.md | Architect | Full-stack architecture |
front-end-spec-tmpl | docs/front-end-spec.md | UX Expert | UI/UX specifications |
story-tmpl | docs/stories/{epic}.{story}.md | SM | Story with complete context |
qa-gate-tmpl | docs/qa/{story}-qa-gate.md | QA | Quality gate decision |
Detailed Documentation: Template Reference
| Team | Agents | Platform | Use Case |
|---|---|---|---|
team-fullstack | 6 planning agents | Web platforms | Full lifecycle planning |
team-ide-minimal | PO, SM, Dev, QA | Desktop IDEs | Development only |
team-no-ui | 5 planning agents | Web platforms | Backend/service planning |
team-all | All 10 agents | Any | Maximum flexibility |
Detailed Documentation: Agent Teams Reference
| Pack ID | Version | Description |
|---|---|---|
bmad-godot-game-dev | 1.0.0 | Godot game development (GDScript & C#) |
bmad-2d-unity-game-dev | 1.0.0 | Unity 2D game development |
bmad-2d-phaser-game-dev | 1.0.0 | Phaser 2D game development |
bmad-creative-writing | 1.0.0 | Creative writing & storytelling |
bmad-infrastructure-devops | 1.0.0 | DevOps & infrastructure |
Detailed Documentation: Expansion Packs Reference
your-project/
├── .bmad-core/ # Core framework
│ ├── agents/ # 10 agent personas
│ ├── tasks/ # 21 core tasks
│ ├── templates/ # 13 document templates
│ ├── workflows/ # 6 workflow definitions
│ ├── checklists/ # 6 quality checklists
│ ├── agent-teams/ # 4 team configurations
│ ├── data/ # Knowledge base files
│ └── core-config.yaml # Project configuration
├── common/ # Shared resources
│ ├── tasks/ # 2 common shared tasks (create-doc, execute-checklist)
│ └── utils/ # Common utilities
├── .{expansion-pack}/ # Optional expansion packs
├── docs/ # Generated documentation
│ ├── prd.md # Product Requirements
│ ├── architecture.md # Architecture document
│ └── stories/ # Story files for development
└── web-bundles/ # Optional web bundles for ChatGPT/Claude/GeminiCore configuration is in .bmad-core/core-config.yaml:
markdownExploder: true # Use markdown-tree-parser for sharding
qa:
qaLocation: docs/qa # QA artifacts directory
prd:
prdFile: docs/prd.md # PRD file path
prdVersion: v4 # Format version
prdSharded: true # Enable sharding
prdShardedLocation: docs/prd # Sharded directory
epicFilePattern: epic-{n}*.md # Epic file naming pattern
architecture:
architectureFile: docs/architecture.md
architectureVersion: v4
architectureSharded: true
architectureShardedLocation: docs/architecture
devLoadAlwaysFiles: # Files dev always loads
- docs/architecture/coding-standards.md
- docs/architecture/tech-stack.md
devStoryLocation: docs/stories # Story directory
slashPrefix: "*" # Command prefixDetailed Documentation: Configuration Reference
| Command | Purpose | Example |
|---|---|---|
install | Install/upgrade BMad | npx bmad-method install |
update | Update installation | npx bmad-method update |
status | Show installation status | npx bmad-method status |
flatten | Flatten codebase to XML | npx bmad-method flatten -i ./src -o out.xml |
list:expansions | List available packs | npx bmad-method list:expansions |
list:agents | List all agents | npx bmad-method list:agents |
Detailed Documentation: CLI Reference
bmad-method can be used programmatically in Node.js applications:
const installer = require('bmad-method/tools/installer/lib/installer');
const WebBuilder = require('bmad-method/tools/builders/web-builder');
const DependencyResolver = require('bmad-method/tools/lib/dependency-resolver');
// Install BMad programmatically
await installer.install({
installType: 'full',
directory: './my-project',
ides: ['claude-code', 'cursor']
});
// Build web bundles
const builder = new WebBuilder({ rootDir: process.cwd() });
await builder.buildAgents();
// Resolve dependencies
const resolver = new DependencyResolver('./my-project');
const agents = await resolver.listAgents();Detailed Documentation: Programmatic API Reference
Desktop IDEs: Cursor, Claude Code, Windsurf, Trae, Roo Code, Kilo Code, Cline, VS Code (GitHub Copilot), Crush, Qwen Code
CLI Tools: iFlow CLI, Auggie CLI (Augment Code), Gemini CLI, Codex CLI
Web Platforms: ChatGPT, Claude, Gemini, Codex Web, OpenCode
All agents follow consistent activation:
.bmad-core/core-config.yaml*helpAll commands use * prefix (e.g., *help, *create-prd, *develop-story).
For detailed examples: See Real-World Scenarios and Edge Cases for comprehensive usage examples.
Phase 1: Planning (Web Platform)
*agent analyst # Create project brief
*agent pm # Create PRD with epics and stories
*agent ux-expert # Create frontend specification
*agent architect # Create full-stack architecture
*agent po # Validate and shard documentsPhase 2: Development (IDE)
*agent sm # Create story files with complete context
*agent dev # Implement stories sequentially
*agent qa # Review and gate implementationPhase 1: Document & Plan
npx bmad-method flatten -i . -o project.xml # Document existing codebase
*agent pm # Create brownfield PRD
*agent architect # Document architecture changes
*agent po # Shard documentsPhase 2: Development
*agent sm # Create brownfield stories
*agent dev # Implement with safety checks
*agent qa # Validate changesLarge documents (PRD, Architecture) are split into manageable chunks:
*agent po
*shard-doc
# Output:
# docs/prd/ with epic-1.md, epic-2.md, etc.
# docs/architecture/ with tech-stack.md, coding-standards.md, etc.SM agent creates stories with complete implementation context:
interface StoryContext {
requirements: 'From sharded PRD epic';
architecture: 'Selective sections (tech-stack, backend-architecture, etc.)';
previousStories: 'Learnings from completed stories';
sourceReferences: '[Source: docs/architecture/tech-stack.md#frameworks]';
tasks: 'Checkbox list with subtasks';
}QA agent makes gate decisions with detailed rationale:
interface QualityGateDecision {
PASS: 'All criteria met, ready for production';
CONCERNS: 'Issues found but not blocking';
FAIL: 'Critical issues, must address';
WAIVED: 'Issues waived by stakeholder';
}Templates use sophisticated markup for AI-driven generation:
interface TemplateMarkup {
placeholders: {
syntax: '{{variable_name}}';
example: '{{project_name}}, {{epic_num}}';
};
llmInstructions: {
syntax: '[[LLM: instructions]]';
visibility: 'Invisible to users, processed by AI';
};
elicitation: {
property: 'elicit: true';
format: 'Mandatory 1-9 options with methods';
};
}Agents load dependencies on-demand:
interface DependencyResolution {
pattern: '{root}/{type}/{name}';
types: ['tasks', 'templates', 'checklists', 'data'];
example: 'create-doc.md → .bmad-core/tasks/create-doc.md';
}