BMAD-METHOD™ is a Universal AI Agent Framework providing two key innovations for AI-assisted delivery:
-
Agentic Planning: Dedicated agents (Analyst, PM, Architect) collaborate with you to create detailed, consistent PRDs and Architecture documents. Through advanced prompt engineering and human-in-the-loop refinement, these planning agents produce comprehensive specifications that go far beyond generic AI task generation.
-
Context-Engineered Development: The Scrum Master agent transforms detailed plans into hyper-detailed development stories containing everything the Dev agent needs - full context, implementation details, and architectural guidance embedded directly in story files. QA provides Test Architect capabilities with risk profiling, test design, and quality gates.
This two-phase approach eliminates both planning inconsistency and context loss - the biggest problems in AI-assisted development. Your Dev agent opens a story file with complete understanding of what to build, how to build it, and why.
Core components live under bmad-core/ (10 agents, 21 tasks, 6 workflows, templates, checklists, knowledge base) and are packaged for IDE and Web UI environments. Expansion packs extend the framework to any domain: game development, creative writing, DevOps, and more.
Install (IDE): npx bmad-method install Web UI: Upload team bundle to Gemini Gem or CustomGPT Key roles: BMad-Master, BMad-Orchestrator, Analyst, PM, Architect, PO, SM, Dev, QA (Test Architect), UX Expert.
Flows
Planning Workflow (Web UI / Powerful IDE Agents)
Structured PRD and Architecture creation with optional research and UX steps.
Core Development Cycle (IDE)
SM drafts detailed stories; Dev implements; QA (Test Architect) reviews with risk profiling and quality gates; loop until done.