bmad-code-org/BMAD-METHOD

What it solves

BMad addresses the tendency of AI coding assistants to make unstated assumptions during implementation. It provides a structured, agile framework for "AI Driven Development" (AiDD) that ensures product and technical decisions remain explicit and are preserved as durable context, preventing the need to re-explain requirements in every chat session.

How it works

BMad integrates with AI coding tools (such as Claude Code or Codex) via plugins and a CLI. It uses a delivery loop consisting of Clarify, Plan, Build, and Learn stages. Depending on the complexity of the task, users can either go straight to implementation for simple changes or employ deeper planning workflows. The system utilizes specialized agent perspectives (Product, Architecture, UX, Dev, and Testing) to guide collaboration and produce briefs and specifications that drive the actual coding process.

Who it’s for

It is designed for developers and software creators who want to use AI to build software—ranging from weekend prototypes to large, established codebases—without losing control over the architectural and product decisions.

Highlights

• Right-sized process: Flexible paths that allow for immediate building of small changes or deep planning for complex initiatives.
• Durable context: Maintains a record of decisions to carry context forward across the development lifecycle.
• Specialized perspectives: Access to dedicated agent roles for architecture, UX, and testing.
• Broad ecosystem: Includes specialized modules for game development, enterprise testing, and creative intelligence.