Workflow Guide

BMAD Method + StacksFinder MCP: Data-Driven Tech Decisions

The BMAD method provides structure for AI-assisted development. The StacksFinder MCP brings deterministic, data-driven tech decisions into that workflow. Here's how to combine them.

10 min read Updated 2026-01-12

Key Takeaways

  • BMAD + MCP = structured, data-driven development
  • Use analyze_tech & list_technologies in Planning (Phase 2)
  • Use compare_techs & recommend_stack in Solutioning (Phase 3)
  • Document decisions with create_blueprint in Implementation
01

What is the BMAD Method?

BMAD ("Build More, Architect Dreams") is an AI-driven agile methodology from bmad-code-org with 21+ specialized agents and 50+ workflows.

The method organizes development into 4 phases. Tech stack decisions happen primarily in phases 2-3, which is where StacksFinder MCP integrates.

BMAD Workflow with MCP Integration Points
Optional
1

Discovery

Brainstorm, research, explore solutions

brainstormresearchproduct-brief
No MCP tools needed yet
Required
2

Planning

Create PRD, tech specs, UX design

prdtech-specux
StacksFinder MCP
list_technologiesanalyze_tech
Explore & evaluate options
Required
3

Solutioning

Architecture, epics, implementation readiness

architectureepics-storiesvalidate-arch
StacksFinder MCP
compare_techsrecommend_stack
Compare & decide
Required
4

Implementation

Sprint planning, story development, code review

sprint-plandev-storycode-review
StacksFinder MCP
create_blueprintget_blueprint
Document & share

Based on the official BMAD-METHOD workflow. See their full diagram.

02

The Tech Decision Problem

LLM Hallucinations

AI models have training cutoffs and can confidently state outdated or incorrect information about frameworks.

Opinion-Based Advice

"Use React" without context. LLMs reflect popular opinions, not data-driven analysis of your constraints.

Non-Reproducible

Ask the same question twice, get different answers. Makes it impossible to audit or justify decisions.

The Solution: Deterministic Scoring via MCP

The StacksFinder MCP gives your AI assistant access to versioned, deterministic scores for 30+ technologies. Same question = same answer. Every score is explainable and auditable.

03

MCP Tools Overview

Quick Install (30 seconds)

Add to your Claude Claude Code, Cursor Cursor, or VS Code VS Code configuration:

{
  "mcpServers": {
    "stacksfinder": {
      "command": "npx",
      "args": ["-y", "stacksfinder-mcp"]
    }
  }
}
No API key needed for free tools
Full installation guide →
list_technologies Free

Get all 30+ technologies organized by category

When to use: Early exploration of what options exist
Example prompt:

"List all available database technologies"

analyze_tech Free

6-dimension scores, strengths, weaknesses, compatible techs

When to use: Deep dive on a specific technology
Example prompt:

"Analyze SvelteKit for a solo developer building an MVP"

compare_techs Free

Side-by-side comparison of 2-4 technologies

When to use: Narrowing down between finalists
Example prompt:

"Compare Next.js, SvelteKit, and Remix for a SaaS product"

recommend_stack Free

Full stack recommendation for any project type

When to use: Getting a complete stack suggestion
Example prompt:

"Recommend a stack for an e-commerce MVP with a $50/month budget"

create_blueprint Pro

Generate a shareable blueprint with AI narrative

When to use: Creating documentation for stakeholders
Example prompt:

"Create a blueprint for our fintech startup stack"

get_blueprint Pro

Fetch an existing blueprint by ID

When to use: Retrieving previously created blueprints
Example prompt:

"Get blueprint bp_abc123"

04

When to Use Each MCP Tool

The key insight: use data tools in phases 2-3 before committing to architecture decisions. Phase 1 focuses on business goals (no tech yet), and Phase 4 is about documenting what you decided.

Phase
Goal
MCP Tools
Example Prompt
1 Discovery
Explore the problem space
None needed
Focus on business goals, not tech
2 Planning
Research options
list_technologies analyze_tech
"List all database options, then analyze PostgreSQL for a 2-person team with EU compliance needs."
3 Solutioning
Compare & decide
compare_techs recommend_stack
"Compare SvelteKit vs Next.js, then recommend a full stack for B2B SaaS with auth and payments."
4 Implementation
Document & share
create_blueprint get_blueprint
"Create a blueprint documenting our stack choices for the investor technical review."
05

Practical Examples

Scenario: Solo Founder Building a SaaS MVP

Constraints: 1 developer, $50/month budget, need to launch in 6 weeks

You

I'm starting a B2B SaaS for invoice management. Solo founder, $50/month budget, need to ship in 6 weeks. Help me choose a tech stack using the StacksFinder tools.

Claude

I'll use the StacksFinder MCP to get data-driven recommendations. Let me start by getting a full stack recommendation for your constraints...

recommend_stack context: "mvp", budget: "low", team_size: 1
Result

Recommended Stack (Score: 89/100)

  • Framework: SvelteKit (DX: 92, solo-dev bonus)
  • Database: PostgreSQL via Supabase (free tier)
  • Auth: Supabase Auth (built-in)
  • Hosting: Vercel (free tier)
You

I've heard good things about Next.js too. How does it compare?

Claude

Let me compare them directly for your use case...

compare_techs ["sveltekit", "nextjs"], context: "mvp"
Comparison Result

SvelteKit vs Next.js for Solo MVP

  • SvelteKit: 89 (simpler mental model, less boilerplate)
  • Next.js: 84 (larger ecosystem, more learning resources)

Verdict: SvelteKit edges out for solo developers prioritizing speed-to-ship.

You

Perfect. Let's document this for my pitch deck.

Claude

I'll create a shareable blueprint with the rationale...

create_blueprint title: "Invoice SaaS MVP Stack"
Blueprint Created

Blueprint ready: bp_inv_saas_2026

Share link: stacksfinder.com/bp/inv_saas_2026

Includes scoring breakdown, rationale, and export to PDF.

Ready to try data-driven stack decisions?

06

Frequently Asked Questions

What is the BMAD method?

BMAD stands for "Build More, Architect Dreams" (or Business-Method-Architecture-Development). It is an AI-driven agile methodology from bmad-code-org with 4 phases: Discovery (optional), Planning, Solutioning, and Implementation. It uses specialized AI agents for each phase.

Why use StacksFinder MCP instead of asking the LLM directly?

LLMs have training cutoff dates and can hallucinate about technology capabilities. StacksFinder provides deterministic, versioned scores based on current data. The same question always produces the same answer, making decisions reproducible and auditable.

Which AI coding assistants support the MCP?

The StacksFinder MCP works with Claude Code, Cursor, VS Code Copilot (with MCP support), Windsurf, and any MCP-compatible client. Installation takes under 2 minutes via npx.

Is the MCP free to use?

Yes, the core tools (list_technologies, analyze_tech, compare_techs, recommend_stack) are free with generous rate limits. Pro features like create_blueprint and API access require a subscription.

Can I use BMAD without AI agents?

Absolutely. BMAD is a methodology that works with or without AI. However, combining it with AI agents and the StacksFinder MCP significantly accelerates the research and decision-making phases.

Where can I learn more about BMAD?

The official BMAD repository is at github.com/bmad-code-org/BMAD-METHOD. It includes 21+ specialized agents, 50+ workflows, and comprehensive documentation. The BMadCode YouTube channel also has tutorials.

Start Making Data-Driven Tech Decisions

Install the StacksFinder MCP in under 2 minutes and bring deterministic scoring to your AI workflow.

Free tier includes list_technologies, analyze_tech, compare_techs, and 1 recommend_stack per day.