Canopy TechOverview

How Canopy works

Canopy asks the architecture questions your future CTO would ask — before you write a line of code. Three tools, each built for a different moment in your build process.

How it works

From idea to ready-to-build brief in three steps.

01

Describe your project

One sentence in Quick Stack, or answer 18 structured questions in the Full Consultation. Either way takes under 5 minutes.

02

AI generates your architecture brief

Your answers go to Claude with a structured prompt. You get a full stack recommendation, monthly cost estimate, compliance flags, and a corporate readiness score — tailored to your specific constraints.

03

Paste the system prompt and start building

The brief includes a ready-to-use vibe coding system prompt. Drop it into Cursor, Claude Code, Windsurf, or your tool of choice and build with the full context already loaded.

Who it's for

Solo builder / indie hacker

Side project, no team, want fast decisions. Quick Stack gives you a direction in 30 seconds. Full Consultation surfaces the traps you'd otherwise hit at 1,000 users.

Early-stage founder

Building to ship. The compliance flags and corporate readiness score mean you won't discover your GDPR exposure or SSO gap after the first enterprise prospect asks about it.

Agency / freelancer

Building for clients who need to understand what they're buying. The architecture brief is a document you can hand off — stack rationale, cost projections, and a system prompt included.

The three tools

Quick Stack~30 seconds
Try it

For: Developers who want a stack direction before committing to a full interview. Good for quick gut-checks, side projects, and hackathons.

You get

  • Six stack cards — Frontend, Backend, Database, Auth, Styling, Deployment
  • One-line rationale per card, tuned to your description
  • A prompt to take the full consultation for more depth
Full Consultation~5 minutes
Try it

For: Founders and builders starting a project they intend to ship. 18 questions across 5 layers surface everything a principal architect would want to know upfront.

You get

  • Full stack table with primary recommendation + alternative per layer
  • Monthly cost estimate including free-tier ceilings and viral-growth scenarios
  • Compliance flags — GDPR, HIPAA, COPPA — triggered automatically from your answers
  • Corporate readiness score across 5 dimensions (auth, data isolation, compliance, observability, API contracts)
  • A ready-to-paste vibe coding system prompt for Cursor, Claude, Windsurf, or your tool of choice
Repo Audit~30 seconds
Try it

For: Builders with an existing project who want an honest read on how production-ready their stack is. Submit a GitHub URL or paste your code directly.

You get

  • Architecture assessment against the same 5-layer framework
  • Specific gaps identified — missing auth, no RLS, unhandled scale risks
  • Prioritised fix list ordered by severity
  • Upgrade recommendations where applicable

How the Full Consultation works

1

Layer 1 — Project Basics

What the app does, who it's for, what users can do, and whether AI is core to the product.

2

Layer 2 — Users & Scale

Expected user volume, whether accounts are needed, and real-time data requirements.

3

Layer 3 — Data & Privacy

What information the app stores, where users are located, and COPPA exposure.

4

Layer 4 — Constraints

Monthly budget, existing vendor accounts, launch timeline, and which AI coding tool you're using.

5

Layer 5 — Future & Scale

Enterprise potential, regulated industry exposure, monetisation plans, and IT approval requirements.

Brief generated

Your answers go to Claude with a structured prompt. The brief streams back in seconds — you can copy, share, download as PDF, or paste the system prompt straight into your AI coding tool.

Sample output

A trimmed excerpt from a real Full Consultation brief.

🏗️ CANOPY TECH — PROJECT BRIEF

Project:Personal finance tracker with budget visualisations  · Mode: Pre-Build

Your Recommended Stack

LayerRecommendation
FrontendNext.js + TypeScript
AuthClerk
DatabaseSupabase (Postgres)
AI EngineClaude Haiku
HostingVercel

⚠️ Flags & Watch-outs

GDPR

You mentioned European users. You'll need a privacy policy and a delete-my-data endpoint before launch. Add a /api/account/deleteroute that wipes all rows associated with the user's ID.

COST

AI is a core feature. At 10,000 users calling your categorisation endpoint daily, you're looking at ~$400/month in inference costs. Add rate limiting per user from day one.

🤖 Your Vibe Coding System Prompt

COPY THIS INTO CURSOR:

You are building a personal finance tracker.

STACK:
- Frontend: Next.js 15 + TypeScript
- Auth: Clerk (free tier, 10K MAU)
- Database: Supabase Postgres with RLS
- AI: Claude claude-haiku-4-5-20251001 for categorisation
- Hosting: Vercel

RULES:
- Every database query must use RLS — never bypass with service key on client
- Rate-limit AI calls to 20/user/day using Upstash Redis
- All money values stored as integers (cents), never floats

NEVER:
- Store raw API keys in the frontend
- Skip input validation on transaction amounts

The full brief also includes an estimated monthly cost breakdown and a 5-dimension corporate readiness score.

What Canopy covers that others don't

Scaffold generators give you files. Canopy gives you a plan — before the files, and after.

🛡️

Compliance flags

GDPR, HIPAA, and COPPA are triggered automatically from your answers — not just if you think to ask. Most tools generate files; Canopy flags the legal risks before you ship.

💸

Cost planning

Monthly cost estimates with free-tier ceilings and viral-growth projections. Knowing the ceiling before you build prevents a nasty surprise at scale.

🏢

Corporate readiness score

Five dimensions: SSO support, row-level security, compliance posture, observability, and API contracts. Surfaces the enterprise blockers that kill deals months later.

🔍

Retroactive audit

Already shipped something? Repo Audit reviews your existing stack against the same framework. VibeKit and scaffold generators only work on greenfield — Canopy works before and after.

🔓

No account, no limits

Fully free, no sign-up, no credit card, no daily cap. Every tool runs on your own session — nothing is stored server-side.

🤖

Tool-agnostic output

The system prompt works in Cursor, Claude Code, Windsurf, Copilot, or any AI coding tool. Use whatever you already have.

Quick StackFull ConsultRepo Audit
Stack recommendation
Compliance flags (GDPR, HIPAA, COPPA)
Monthly cost estimate
Corporate readiness score
Vibe coding system prompt
Existing codebase review
No account required

FAQ

How is Canopy different from VibeKit or other scaffold generators?

Scaffold generators give you files — project structure, setup instructions, markdown for your AI agent. Canopy gives you a plan: the right stack for your constraints, compliance risks to watch, a cost model, and a readiness score. Use both: plan with Canopy, then generate your scaffold. They're not competing — they cover different moments in the build process.

Do you store my answers or the generated brief?

No. Answers are sent to the AI and immediately discarded server-side. The brief lives in your browser's session storage and is gone when you close the tab. Nothing is logged to a database.

How accurate are the recommendations?

Opinionated, not authoritative. Canopy gives you a strong starting point based on your specific constraints — scale, budget, compliance, existing accounts. Treat it as a senior engineer's first-pass recommendation, not a contract.

Who is this built for?

Vibe coders, indie hackers, and early-stage founders who are comfortable building but want to avoid classic architecture mistakes. You don't need to be technical to use it — the questions are designed in plain English.

Is it free?

Yes, entirely free. No account required.

Can I share the brief with my team?

Yes. The Full Consultation brief has a Share button that generates a URL with the brief encoded in the hash — no server involved. You can also download it as a .txt file or export to PDF.

Five minutes of planning now saves days — sometimes months — of rework later.