Make Me Unicorn
minjikim89/make-me-unicorn
Checklist de lançamento SaaS com validação contra threads reais do HN e Reddit
Ver no GitHubMCP server que funciona como sistema operacional e checklist de lançamento para construtores solo de SaaS. Valida código gerado por IA contra problemas reais, gates de lançamento (Problem Fit, Build Fit, Revenue Fit, etc.) e fornece dashboards de progresso iterativos.
Instalação
npx convez-ai minjikim89/make-me-unicorn1 comando no terminal: detecta o formato do repositório (marketplace ou plugin avulso) e instala via Claude Code. Requer Node 18+ e o Claude Code instalado.
Métricas
Casos de uso
- Validar ideias de startup antes de investir tempo em desenvolvimento
- Auditar projetos gerados por IA para gaps de produção e confiabilidade
- Rastrear progresso de lançamento através de gates estruturados (M0-M5)
- Gerar badges e scorecards de prontidão para marketing inicial
Veredito

Ferramenta pragmática que preenche uma lacuna real para solo founders: transforma sabedoria coletiva (HN, Reddit) em guardrails automáticos durante desenvolvimento. Abordagem modular e CLI limpa favorecem integração com fluxos Claude Code existentes, embora demande curva de aprendizado inicial nos 6 gates de lançamento.
README
A tradução para o português deste README ainda está na fila.
Make Me Unicorn
Catch what AI-generated SaaS code misses — before it costs you users, money, or trust.
The open-source launch checklist and operating system for solo builders.
English · 🇰🇷 한국어 · 🇯🇵 日本語 · 🇨🇳 简体中文 · 🇪🇸 Español
60-Second Demo
pip install make-me-unicorn # or zero-install: uvx make-me-unicorn
cd your-project
mmu init && mmu scan && mmu
MAKE ME UNICORN - STATUS DASHBOARD
.--*--.
/ *v* \
| |
\ ___ /
'---'
Stage: HATCHING ######.............. 22% (124/551)
LAUNCH GATES (16/26)
M0 Problem Fit ################ 4/4 PASS
M1 Build Fit ################ 5/5 PASS
M2 Revenue Fit ############.... 3/4 OPEN
M3 Trust Fit ################ 4/4 PASS
M4 Growth Fit ########........ 2/4 OPEN
M5 Scale Fit ####............ 1/5 OPEN
Your unicorn evolves as you build: Egg → Hatching → Foal → Young → Unicorn → Legendary.
Add a Badge to Your README
Show your launch readiness to the world. One command:
mmu badge # get markdown badge
mmu badge --format svg -o badge.svg # save as SVG file
mmu badge --clipboard # copy to clipboard
Then paste in your README:
Every badge links back here. Every README becomes a growth channel.
Share Your Score
mmu share # print shareable score card
mmu share --clipboard # copy to clipboard (macOS)
┌─────────────────────────────────────────────┐
│ Make Me Unicorn — Launch Readiness │
│ │
│ Score: 68% Stage: YOUNG UNICORN │
│ │
│ M0 Problem Fit ████████████████ PASS │
│ M1 Build Fit ████████████████ PASS │
│ M2 Revenue Fit ██████████░░░░░░ OPEN │
│ M3 Trust Fit ████████████████ PASS │
│ M4 Growth Fit ████████░░░░░░░░ OPEN │
│ M5 Scale Fit ████░░░░░░░░░░░░ OPEN │
│ │
│ Stack: Next.js · Stripe · SSR │
│ pip install make-me-unicorn │
│ github.com/minjikim89/make-me-unicorn │
│ #MakeMeUnicorn │
└─────────────────────────────────────────────┘
Check your SaaS launch readiness:
https://github.com/minjikim89/make-me-unicorn
The Problem
You code with AI. You ship fast. But then:
| What goes wrong | What it costs you |
|---|---|
| Forget password reset while building login | Users locked out on day 1 |
| Skip webhook signature verification | Attackers replay payment events |
| Launch without OG tags | Every shared link looks broken |
| Lose context between AI sessions | Re-explain your project from scratch |
| No refund policy | First dispute = frozen Stripe account |
You're not failing at coding. You're failing at tracking what matters.
What MMU Covers
Building
- Frontend (responsive, a11y, forms)
- Backend (API, DB, queues)
- Auth (login, reset, OAuth, sessions)
- Billing (Stripe, webhooks, refunds)
- Testing (unit, E2E, agent safety)
Launching
- SEO (OG tags, sitemap, meta)
- Legal (privacy, terms, GDPR)
- Security (CORS, rate limits, secrets)
- Performance (caching, lazy load)
- CI/CD (pipeline, rollback plan)
Operating
- Monitoring (errors, uptime, alerts)
- Analytics (funnel, retention, events)
- Email (transactional, templates)
- Accessibility (WCAG, keyboard nav)
- Reliability (backup, incident plan)
670+ items. 15 categories. 6 launch gates. Zero guesswork.
How Is MMU Different?
| Scope | Lives where | Agent-native | Tracks progress | |
|---|---|---|---|---|
| MMU | Full SaaS: code + billing + legal + growth + ops | Your repo, plain markdown | ✅ Claude plugin + MCP server | ✅ Score, gates, evolving dashboard |
| Lighthouse | Frontend performance/a11y/SEO only | Browser/CI | ❌ | Per-run score, no project memory |
| Vendor launch checklists (Vercel, Stripe docs) | One vendor's slice | Their docs site | ❌ | ❌ Read-only |
| Static checklist repos / Notion templates | Varies, usually unverifiable | Copy-paste | ❌ | ❌ Manual, goes stale |
| AI code reviewers (CodeRabbit etc.) | Code diffs | PR workflow | Partially | Per-PR, no launch view |
MMU is not a linter and not a docs site — it's the operating layer between your AI coding sessions and an actual launch: a verifiable checklist your CLI, CI, and AI agents all read and update.
Vibe Check Your AI-Generated Code
45% of AI-generated code ships with vulnerabilities. One command scans for the gaps AI assistants miss most:
mmu vibecheck
Vibe check — what AI-generated code usually misses
✗ (P0) secrets: possible hardcoded secrets in 1 file(s): Stripe live secret key
✗ (P0) webhook-signature: webhook handlers found but no signature verification markers
↳ Verify provider signatures or attackers can forge payment events.
✗ (P0) password-reset: auth code detected but no password reset flow markers
↳ The #1 day-one lockout: users who can log in but can never get back in.
⚠ (P1) rate-limiting: server framework detected but no rate limiting markers
Vibe check result: 3 launch-blocking issue(s), 1 warning(s)
Checks: hardcoded secrets · unignored .env · webhook signature + idempotency · password reset flow · f-string SQL · rate limiting · wildcard CORS · DEBUG = True · error monitoring. P0 findings exit non-zero, so it drops straight into CI.
Personalize Your Checklist
Not every project needs billing or i18n. MMU adapts:
mmu init # generates .mmu/config.toml
[features]
billing = false # no Stripe? billing items won't count against you
i18n = false
[architecture]
framework = "nextjs"
Your score reflects only what applies to your project. mmu status --why shows the breakdown — like Lighthouse, but for SaaS readiness.
6 Launch Gates
Phase exits. Don't skip ahead.
M0 Problem Fit → Do you know WHO and WHY?
M1 Build Fit → Does the core product work end-to-end?
M2 Revenue Fit → Can someone pay you? And get a refund?
M3 Trust Fit → Privacy policy? Support path? Logging?
M4 Growth Fit → Will links look right? Can people find you?
M5 Scale Fit → What happens at 3am?
Run mmu gate --stage M0 to verify.
12 Operating Modes
One mode per session. Each loads only the docs you need — prevents the #1 problem with AI coding: context overload.
mmu start --mode backend # loads: architecture.md, sprint, ADR logs
mmu start --mode billing # loads: pricing.md, billing checklist, compliance
mmu start --mode growth # loads: SEO checklist, metrics
AI Integration (Optional)
MMU works without AI. But with Claude, it gets powerful:
pip install make-me-unicorn[llm]
export ANTHROPIC_API_KEY=sk-ant-...
| Command | What happens |
|---|---|
mmu init --interactive | Answer 5 questions → Claude writes your strategy, product, pricing, architecture, and UX docs |
mmu start --mode X --agent | Auto-formats session context for Claude Code or any LLM |
mmu doctor --deep | Claude reads your code, flags mismatches, security gaps, blind spots |
mmu generate strategy | Generates or updates core docs based on current project state |
Core CLI stays zero-dependency. AI features degrade gracefully.
Use as a Claude Skill
MMU is also packaged as a Claude Code plugin and Anthropic Agent Skill, so any Claude-based agent (Claude Code, Claude Desktop, or any tool that supports the Agent Skills spec — including OpenAI Codex CLI) can auto-invoke MMU when the user mentions a startup idea, validation, launch checklist, or Product Hunt prep.
# In Claude Code:
/plugin marketplace add minjikim89/make-me-unicorn
/plugin install make-me-unicorn
The skill auto-loads only the blueprint(s) relevant to the conversation (progressive disclosure), so it stays cheap on context.
MCP Server Mode
MMU also runs as an MCP (Model Context Protocol) server, so any MCP-compatible agent (Claude Code, Claude Desktop, Cursor, Gemini CLI) can call MMU's blueprints and templates as native tools.
pip install make-me-unicorn[mcp]
mmu serve-mcp # stdio transport (default)
mmu serve-mcp --transport sse # SSE transport
Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"mmu": {
"command": "mmu",
"args": ["serve-mcp", "--root", "/path/to/cloned/make-me-unicorn"]
}
}
}
Tools exposed:
mmu_list_blueprints— list 17 blueprints (15 core + 2 industry)mmu_get_blueprint(name)— fetch full blueprint markdownmmu_list_idea_templates— list start/close/ADR prompts + Product Hunt kitmmu_validate_idea(idea)— validate against real HN + Reddit threads: verdict, sentiment, competitors, top threads (free mode, no API keys; needs the[validate]extra)
Validate an Idea
Pull real signal from HN + Reddit before you build:
pip install make-me-unicorn[validate]
mmu validate "AI tutor for kids" --limit 30
Default mode is free — no API keys, no paid calls. Public Reddit + HN Algolia search, local VADER sentiment, capitalized-token competitor surface. Saves a markdown report to reports/validate/<slug>-<hash>.md.
For a 1-page validation verdict synthesized from the threads:
mmu validate "AI tutor for kids" --llm
# Prompts for cost confirmation (~$0.05-0.20). Use -y to skip.
--llm is opt-in — the default flow never calls the Anthropic API.
Full Command Reference
mmu # status dashboard
mmu init # bootstrap project
mmu init --interactive # LLM-guided setup (5 questions → 5 docs)
mmu scan # auto-detect tech stack
mmu status --why # score breakdown
mmu next # prioritized next actions
mmu show frontend # drill into any category
mmu check frontend 3 # mark item done
mmu gate --stage M0 # verify gate readiness
mmu doctor # guardrail health checks
mmu doctor --deep # LLM-powered semantic review
mmu vibecheck # scan for AI-generated code blind spots (secrets, webhooks, …)
mmu share # shareable score card
mmu badge # README badge (markdown/svg/html)
mmu start --mode backend # start focused session
mmu close # end session with structured memory
Who This Is For
| You are... | MMU helps you... |
|---|---|
| A founder coding with AI | Stop re-explaining your project. Keep context across tools. |
| A frontend developer | Know exactly what to build: auth flows, error states, OG tags. |
| A product manager | Structured PRD, pricing strategy, launch checklist — all in markdown. |
| A fullstack builder | Track everything in one place. Nothing slips through. |
Example: TaskNote
A fully filled-out example:
examples/filled/tasknote/
├── docs/core/strategy.md ← ICP, value prop, competitors
├── docs/core/product.md ← MVP scope, user journey, P0/P1
├── docs/core/pricing.md ← Free/Pro/Team, billing rules
├── docs/core/architecture.md ← Next.js + FastAPI + Postgres
├── docs/adr/001_billing_provider_choice.md ← Why Stripe?
└── current_sprint.md ← This week's 3 goals
Requirements
- Python
3.10+ - Core CLI: zero external dependencies
- AI features:
pip install make-me-unicorn[llm]
CI Guardrails
mmu doctor runs on every PR. mmu gate runs for stages listed in docs/ops/gate_targets.txt.
Contributing
See CONTRIBUTING.md.
License
MIT. See LICENSE.
mcp-name: io.github.minjikim89/make-me-unicorn