bmad-code-org/BMAD-METHOD: Breakthrough Method for Agile Ai Driven Development

Breakthrough Method for Agile Ai Driven Development - bmad-code-org/BMAD-METHOD

Summary

BMAD-METHOD is a Python-based workflow and toolset for AI-driven software development that keeps product and technical decisions explicit and reusable across sessions. It scales from small changes to large projects, on both new code and inherited codebases, and runs inside AI coding assistants that support skills, such as Claude Code and Codex.

What it is useful for

The README targets a recurring problem: coding assistants are good at implementation but tend to bake unstated assumptions directly into code. BMAD counters that by producing persistent briefs, specifications, and architecture that travel with the work, so context does not have to be re-explained in every chat.

The process is right-sized. A clear, small change can go straight to build, while larger work gets deeper planning and brings in specialist perspectives covering product, architecture, UX, development, and testing. The documented delivery loop runs Clarify, then Plan, then Build and verify, with a Learn phase feeding adjustments back into Plan. Engineers can start a new project or pick up an existing codebase; in the latter case, the workflow begins by establishing verified context on what is actually there before changing it. Add-on modules cover enterprise testing, game development, and creative work.

How engineers can use it

The README states that installation requires an AI coding tool that supports skills and uv for the Python scripts. Three documented install routes are given.

Skills CLI, in the project directory:

npx skills add bmad-code-org/BMAD-METHOD

Claude Code plugin, inside Claude Code:

/plugin marketplace add bmad-code-org/bmad-plugins

Codex plugin, from the terminal:

codex plugin marketplace add bmad-code-org/bmad-plugins

After installation, open the coding tool in the project and ask the bmad skill to run bmad setup. To make a change, invoke bmad-build with a description of what you want modified. bmad status reports versions and suggests the next step; bmad setup also handles updates by running npx skills update and cleaning up renamed or removed skills.

Documented constraints: setup depends on a skills-compatible AI coding tool plus uv, so the workflow will not run inside arbitrary editors. The GitHub metadata shows license "NOASSERTION" even though the README references an MIT LICENSE file, so verify the LICENSE file before vendoring. For deeper walkthroughs, the README points to docs.bmad-method.org, including pages on building a first change and choosing a planning path.