Multi-agent BMAD framework for Claude Code — SPEC → DESIGN → BUILD → QA in one kit

Checking prerequisites...

✔ Node.js (≥18) v24.10.0

✔ Claude Code CLI @anthropic-ai/claude-code

✔ Git v2.47.1

✔ All prerequisites met

? What would you like to do?

› /init-kit — Bootstrap AGENTS.md for a new project (~8 questions)

/create-feature — Run the full BMAD pipeline (SPEC → DESIGN → BUILD → QA)

/create-spec — BA only — turn a requirement into SPEC.md

BMAD multi-agent kit for Claude Code — one command scaffolds the agents, commands and skills that drive a feature from SPEC to deploy.

Quick Start

0. Install Claude Code CLI

Requires Node.js ≥ 18.

npm install -g @anthropic-ai/claude-code

claude --version

Run claude once to sign in (Anthropic account or org API key). The IDE extension (VS Code / JetBrains) or Desktop app also works.

1. Init the kit into your project

npm i agentcrew-kit

npx agentcrew-kit init

2. Add source repos

The kit makes no assumption about repo count/naming — 1 repo (monorepo) or N repos all work. Recommended layout:

<project-name>/

├── CLAUDE.md · POLICIES.md · AGENTS.md · .claude/ · template/ ← kit

├── <project-name>-repository/ ← place cloned repos here

│ ├── <backend-repo>/ <web-repo>/ <mobile-repo>/

└── <project-name>-docs/ ← (optional) dedicated docs repo

3. Init

/init-kit

init-agent asks ~8 questions and fills AGENTS.md with your actual repo paths. Verify the Ecosystem/Actors tables afterwards; re-run to add missing pieces.

4. First feature

/create-feature <feature> [description] # Planning (BA → Design → Tasks, stops at gate)

/create-feature <feature> build # Build after approval (Dev → QA → QC)

Or in natural language: "act as BA, create SPEC for <requirement>". The BMAD pipeline drives itself via the "Next step" at the end of each output. PM steps (/create-plan, /create-backlog) run separately when timeline/assignees are needed.

5. MkDocs site to browse docs

Ask Claude Code:

"Set up MkDocs so I can browse the feature docs at localhost:8000."