Agent-Driven Product Ops
A progressive disclosure walkthrough for builders who want a unified, multi-agent pipeline that connects PRDs, architecture, development, and QA across Claude, Cursor, and web chat surfaces.
You can treat this guide like an expanding ladder. Start with the essentials and open each tier as you’re ready for deeper configuration. The entire setup took me under 20 minutes once I knew the flow.
Core Command
npx bmad-method install
Bundles Path
/Users/chriscruz/Dropbox/Github/BMAD/web-bundles
Agents Activated
pm, po, sm, architect, dev, qa, ux-expert, analyst, bmad-master, bmad-orchestrator
Layer 1
BMad is a modular, opinionated framework built for AI-powered product development. I lean on it for weekly sprints that cycle through PRDs, architecture, sharded stories, and QA gates—exactly the kind of iterative flow modern AI teams live in.
Because it already understands multi-agent orchestration, the method snaps into platforms like Claude, Cursor, and ChatGPT without duct tape. Think of it as a playbook plus automation spine that keeps product, dev, architecture, and QA in sync.
▶️ Watch the BMad Method demoLayer 2
Kick off the interactive wizard with:
npx bmad-method installThis command seeds the .bmad-core directory, prompts for key options, and scaffolds rules you’ll reuse everywhere.
Enter the full path to your project directory where BMad should be installed:
I pointed it at the actual repo I open in Cursor and Claude Code:
/Users/chriscruz/Dropbox/Github/BMADPro tip: choose your main project root so agent rules, docs, and bundles live exactly where your IDE expects them.
Will the PRD (Product Requirements Document) be sharded into multiple files?
BMad is flexible—stick to a single PRD if you’re small, or shard into modular files for enterprise-scale automation. I weigh project size and iteration speed; both choices plug into the same agent loop.
Which IDE(s) do you want to configure?
Navigate with the arrow keys, toggle with SPACE, and confirm with ENTER. I selected Cursor and keep an eye out for native Claude Code support. If you miss one, hit N at the confirmation prompt and rerun.
Would you like to include pre-built web bundles? (standalone files for ChatGPT, Claude, Gemini)
I say YES every time. The installer generates ready-to-upload bundles so you can activate BMad agents in web chat. Drop the bundle into Claude or ChatGPT, follow the instructions, and your agents spin up instantly.
Enter directory for web bundles: (/Users/chriscruz/Dropbox/Github/BMAD/web-bundles)
I take the default path. Keeping bundles next to the repo makes versioning and sharing painless.
Layer 3
/Users/chriscruz/Dropbox/Github/BMAD/web-bundles.@pm, @architect, and more..bmad-core directory with workflow logic, prompt scaffolds, and documentation.Layer 4
The docs live at docs/user-guide.md and .bmad-core/user-guide.md. They outline agent commands, sharding patterns, and quality gates.
@pm, @architect, or their slash-command equivalents.Define your PRD, shard user stories, ship code with dev + architect guardrails, and let QA agents close the loop. Everything is logged, replayable, and ready for the next sprint.
Layer 5
TL;DR: BMad is lightweight to install, heavy on leverage. Once your agents are in play, you can build, iterate, and ship like an AI-native team.
Kick off your next product sprint with BMad at the core. The orchestration spine keeps PRDs, architecture, dev, and QA humming in unison across every AI surface you use.