Who BMAD-METHOD is for#
Developers building AI-assisted projects with architectural complexity
BMAD-METHOD's planning flows produce architecture decision records, briefs, and specs that stay as context through implementation. For a project where the AI's architectural choices need to align with an existing system or a deliberate design, the structured planning step surfaces those decisions before any code is written.
Skip if:
Simple scripts, one-file utilities, or projects where the full scope is clear from the start. Adding planning overhead to genuinely small tasks slows delivery without improving decision quality.
Teams onboarding onto an inherited codebase
The existing-codebase flow reads the current code and produces an explicit context document before any new work begins. This gives the AI a grounded starting point and prevents it from making changes that contradict design decisions already baked into the system.
Skip if:
Green-field projects with no existing codebase context to establish. The onboarding flow adds no value when there is nothing to onboard from.
Solo developers working across multiple sessions on a longer project
Context loss between chat sessions is the core problem BMAD-METHOD addresses for solo developers. Decisions, constraints, and trade-offs captured in planning documents remain available at the start of every new session, so a project that spans weeks does not require re-explaining its architecture each time.
Skip if:
Developers who keep all context in their head or in code comments and find structured briefs an overhead. The framework adds value in proportion to how much a project benefits from explicit decision records.
Teams evaluating Amazon Kiro alternatives
Teams interested in spec-driven AI development but who want to avoid AWS lock-in or a cloud-managed environment will find BMAD-METHOD covers the same spec-driven delivery workflow. It runs locally, supports multiple AI coding tools, and has no cloud service dependency.
Skip if:
Teams already invested in the AWS ecosystem who need Kiro's native AWS service tooling. BMAD-METHOD does not replicate those integrations.
The problem it solves#
AI coding tools are effective at generating code but treat every conversation as a blank slate. Architecture decisions made last sprint, product trade-offs discussed with the team last week, and the specific constraints of the codebase you inherited are invisible to the assistant unless you re-explain them in every session.
The pain deepens on existing codebases. When a developer asks an AI tool to add a feature, the assistant fills in the gaps with assumptions. Those assumptions become code, and the code becomes the implied design. Nobody catches the drift until a decision that seemed minor at the time turns out to have foreclosed a significant architectural path.
How it solves it#
Right-Sized Planning Paths
Three paths are available: go straight to build for clear, bounded changes; run a lightweight planning pass for moderate-complexity features; or run the full product-architecture-engineering planning sequence for new initiatives. The process sizes itself based on the scope of what you ask the bmad-build command to do.
Durable Decision Context
Briefs and specifications written during planning are stored as project files and referenced by the implementation steps that follow. The AI does not re-derive decisions already recorded. This means a question answered during an architecture session stays answered through the build, review, and retro steps.
Specialized Agent Perspectives
Separate agent roles cover product management, architecture, UX, engineering, and testing. Each perspective contributes to the stage where it matters: product and architecture roles inform the plan, engineering and testing roles own the build and verification steps. Running multiple roles in sequence surfaces the trade-offs a single-role session would miss.
New and Existing Codebase Support
A dedicated flow establishes verified context on an inherited codebase before any new work begins. The process reads the existing code, identifies what it does and how it is structured, and surfaces that as explicit context. Changes then build from what is actually there, not from what the assistant assumes.
Multi-Tool Plugin Support
The v7 preview installs via npx skills add bmad-code-org/BMAD-METHOD --skill '*' and includes a plugin marketplace for Claude Code (/plugin marketplace add bmad-code-org/bmad-plugins) and Codex. The stable v6.12 release works in any AI coding tool via the npm installer without tool-specific plugins.
Strengths and trade-offs#
Strengths
- Runs Inside Your Existing AI ToolBMAD-METHOD is not a separate IDE or cloud service. It installs as skills and workflows into whatever AI coding tool you already use. There is no new interface to learn, no account to create, and no service that handles your code outside your own environment.
- No Paywalled Workflows or Gated CommunityThe framework is MIT licensed and the full delivery workflow is available at no cost. There are no premium tiers, no workflow features locked to a paid plan, and no Discord roles gating community support. The ecosystem modules (BMad Builder, BMad Loop, and others) are also open source.
- Right-Sizing Avoids Over-ProcessNot every change needs a full planning sequence. BMAD-METHOD's delivery loop has explicit entry points: a small, clear change goes directly to the build step, skipping planning entirely. The process adds overhead only where the work is complex enough to need it.
- Large, Active Community and Ongoing DevelopmentThe repository has over 54,000 GitHub stars and was last pushed October 2026, with v7 in active preview alpha. A Discord community, a YouTube channel at @BMadCode with tutorials, and 78 open issues on a project at this scale indicate active maintenance and accessible support.
Trade-offs
- -V7 Preview Has Breaking Changes Between UpdatesV7 is in preview alpha with documented breaking changes between updates. The v6.12 stable release is recommended when stability matters. Teams running v7 need to be able to pivot as the install path, skills, and documentation are still changing; locking dependencies is not yet viable on the preview branch.
- -Requires Python 3.10+, Node.js 20.12+, and uvBoth Python 3.10+ and Node.js 20.12+ plus the uv Python package manager are prerequisites for BMad setup and scripts. On machines where one of those runtimes is not installed, setup is a multi-step process before the framework itself is ready to use.
- -No GUI, Dashboard, or Visual Project TrackerBMAD-METHOD has no web interface, dashboard, or project management UI. All interaction happens through your AI coding tool's skill or plugin system. Teams expecting a visual collaboration layer or a hosted project board will not find one here.
BMAD-METHOD vs alternatives#
BMAD-METHOD vs Amazon Kiro
Both tools apply spec-driven development to AI-assisted software delivery, generating explicit specifications and plans before implementation. The practical difference is in how they run: BMAD-METHOD installs into your local project directory and works inside any AI coding tool you already use, while Amazon Kiro is a proprietary IDE that integrates with the AWS ecosystem.
| Feature | BMAD-METHOD | Amazon Kiro |
|---|---|---|
| License | MIT with trademark notice | Proprietary |
| Self-hosting | Yes, local install | No |
| Cloud dependency | None | AWS ecosystem |
| AI tool support | Claude Code, Codex, and others | Kiro's built-in agent |
| Cost | Free | Commercial |
BMAD-METHOD is the better choice for teams that want spec-driven AI development without AWS lock-in, who prefer working in their existing AI coding tool, or who have compliance or data-privacy reasons to keep all code and decisions on their own infrastructure. The MIT license means there are no licensing costs regardless of team size or commercial use.
Amazon Kiro is worth considering for teams already invested in the AWS ecosystem who benefit from Kiro's native AWS service integrations and want a fully managed IDE environment rather than a framework they install into their own setup.
Quick start#
Install into your project directory using the npm installer; Node.js 20.12+, Python 3.10+, and the uv package manager are required.
```bash
# Stable release (v6.12)
npx bmad-method install
# V7 preview alpha (skills-based)
npx skills add bmad-code-org/BMAD-METHOD --skill '*'
```What it's built on#
- Languages
- JavaScriptPython
FAQ#
Is BMAD-METHOD free to use commercially?
Yes. The code is MIT licensed, so you can use it in commercial projects, modify it, distribute it, and build products on it without licensing fees. The trademarks BMad, BMad Method, and BMad Core are reserved separately, meaning you cannot publish a product under those names without permission, but the framework itself carries no commercial restrictions.
Does BMAD-METHOD work with Claude Code and Codex?
Yes. The v7 preview adds native plugin marketplace support for both Claude Code (via /plugin marketplace add bmad-code-org/bmad-plugins) and Codex. The stable v6.12 release works with any AI coding tool via the npm installer, without tool-specific plugins. V7 also supports skills-based install via npx skills add.
What is the difference between v6.12 and v7?
V6.12 is the stable release, recommended when you need a consistent setup that will not break between updates. V7 is a preview alpha with skills-based install and plugin marketplace support; its install path, skills, and documentation are still changing. The v7 release target was October 2026. Both versions require Node.js 20.12+, Python 3.10+, and the uv package manager.
How is BMAD-METHOD different from using an AI coding assistant directly?
Standard AI coding assistants start each session with no memory of previous decisions. BMAD-METHOD adds structured planning workflows that produce persistent decision documents (briefs, specs, architecture notes) that carry forward as context. An AI working within the BMad framework builds from decisions the team explicitly made, rather than filling in gaps with assumptions that may contradict earlier choices.
Can BMAD-METHOD be used on existing codebases?
Yes. There is a dedicated flow for establishing verified context on an inherited or existing codebase before new work begins. The process reads the actual code, builds an explicit context document, and then routes work through the same planning and build steps as a new project. The README links directly to the existing-codebase start guide in the documentation.
Similar open-source tools#
OpenSpec
72,368 GitHub starsSpec-driven development for AI coding assistants
Alternative to
Amazon Kiro
rakazo
4,431 GitHub starsAI teammates you own: your keys, your model, your machine.
Alternative to
Grok Bot
Athena-Public
615 GitHub starsPersistent memory for your AI across IDEs
Alternative to
Pieces for Developers
Pstack Claude
2,690 GitHub starsStructured coding workflows for Claude Code, Codex, and Pi
Recall
757 GitHub starsLocal project memory for Claude Code, generated entirely offline.
Alternative to
ContextPool
Gstack
136,321 GitHub starsTurn Claude Code into a virtual engineering team with 23 skills