🛡️ The Architecture Guardian for AI-Assisted Code

Stop AI from turning your repo into
architectural spaghetti.

AI coding assistants write 300 lines in seconds — and quietly bypass all your service boundaries. RepoGuard auto-generates strict rules, runs real-time git diff audits, and scores PRs from A+ to F.

✨ Cursor (.cursorrules)

🤖 Claude Code (CLAUDE.md)

🌊 Windsurf (.windsurfrules)

🐍 Python (FastAPI/Django)

🐹 Go (Clean Architecture)

⚡ Next.js & NestJS Ready

View Pricing & Plans →

bash — user@workspace: ~/my-fullstack-app

$ npx repoguard-rules audit

🔍 Scanning entire codebase for architectural violations...

====================================================

ARCHITECTURAL HEALTH DASHBOARD

====================================================

Files Audited: 42

Total Violations: 0

🛡️ Health Score: 100/100 [Grade: A+]

✨ Flawless architecture! Zero drift detected across all files.

----------------------------------------------------

💡 Need automated PR blocking & custom rules for your team?

👉 Upgrade to RepoGuard Pro ($12/mo) or Team ($39/mo):

https://taylormatematica-beep.github.io/repoguard/#pricing

----------------------------------------------------

Interactive Sandbox

Test Your Code in the Browser

Paste any code snippet or select a stack preset to see RepoGuard's 12-rule engine detect architectural drift in ~12ms.

Client Engine (12ms)

Zero server latency • Runs in-browser

Before vs After

The "AI Code Rot" Problem

AI models lack holistic architecture memory. Without RepoGuard, here is how tech debt creeps in silently.

❌ Without RepoGuard (AI Spaghetti) FAIL · 58/100 (D)

// components/UserProfile.tsx
export default function UserProfile({ id }: any) {
  // 🚨 Violation: Direct Prisma call in UI component!
  const user = await prisma.user.findUnique({ where: { id } });

  // 🚨 Violation: Direct browser global in SSR!
  const token = window.localStorage.getItem('token');

  // 🚨 Violation: Hardcoded secret leak!
  const secretKey = "sk_live_98127391827391823";

  console.log("Found user:", user);
  return <div>{user.name}</div>;
}
          

✅ Guarded by RepoGuard PASS · 100/100 (A+)

// components/UserProfile.tsx
import { UserService } from '@/services/user.service';
interface UserProfileProps { id: string; }

export default async function UserProfile({ id }: UserProfileProps) {
  // ✓ Clean Architecture: Encapsulated service layer
  const user = await UserService.getUserProfile(id);

  // ✓ Safe secret injection
  const secretKey = process.env.APP_SECRET_KEY;

  return <div>{user.name}</div>;
}
          

⚡ NEW IN v1.6.1

Native Model Context Protocol (MCP) Server

Connect RepoGuard directly to Cursor, Claude Desktop, and Claude Code CLI via local stdio JSON-RPC 2.0. No server hosting or API keys needed.

⚡ Cursor AI

~/.cursor/mcp.json

{
  "mcpServers": {
    "repoguard": {
      "command": "npx",
      "args": ["-y", "[email protected]", "mcp"]
    }
  }
}

🤖 Claude Desktop

claude_desktop_config.json

{
  "mcpServers": {
    "repoguard": {
      "command": "npx",
      "args": ["-y", "[email protected]", "mcp"]
    }
  }
}

Rules Engine

12 Strict Architectural Guardrails

Zero config needed. Multi-language architectural defense across TypeScript, Python, and Golang.

RULE-01 Error

No Direct DB in UI

Blocks raw ORM/Prisma/SQL queries inside React components and API controllers. Enforces dedicated service layers.

RULE-PY-01 Python

FastAPI Layer Separation

Prohibits direct ORM queries and raw database commits (db.commit()) inside FastAPI route handlers.

RULE-GO-01 Golang

Go Clean Architecture

Prohibits raw database/GORM mutations inside Gin, Fiber, or Echo HTTP handlers. Enforces repository pattern.

RULE-GO-02 Golang

No Silenced Go Errors

Catches errors discarded with blank identifiers (_ = err), preventing AI from suppressing compiler strictness.

RULE-02 Critical

Secret Leak Blocker

Catches hardcoded API keys, tokens, and credentials before they get committed to GitHub.

RULE-09 Critical

Client-Side Secret Exposure

Detects sensitive credentials accidentally prefixed with NEXT_PUBLIC_ or VITE_ bundled into browser JS.

RULE-03 Warning

No Lazy ": any"

Prevents AI models from escaping type safety by throwing : any or as any to pass compilation.

RULE-04 Info

Logger Hygiene

Flags stray console.log calls and enforces centralized, structured production loggers.

RULE-05 Error

SSR Hydration Guard

Prevents Next.js Server Components from calling window or localStorage directly.

RULE-06 Critical

SQL Injection Hazard

Detects dangerous unescaped string interpolations in raw SQL queries and requires parameterized input.

RULE-07 Warning

Unvalidated Payloads

Requires incoming API route payloads to be parsed through Zod schemas before being used.

RULE-08 Info

Anti-Duplicate Utils

Stops AI assistants from re-inventing standard helpers (date formatters, validators) that already exist in /utils.

GitHub Action Bot

Automated PR Review Bot (v1.6)

Every Pull Request gets audited inline with an interactive scorecard and exact code suggestions.

🛡️

repoguard-bot bot reviewed 2 minutes ago

🛡️ RepoGuard Architectural Review

ACTION REQUIRED

Metric	Result	Target
Health Score	58/100 (Grade: D)	≥ 85/100
Critical Hazards	1	0
Layer Violations	1	0

❌ [RULE-01] Layering Violation in src/app/checkout/page.tsx:14

Direct Prisma query detected in UI controller. Move logic to /services/checkout.service.ts.

Transparent Pricing

Built for Solo Hackers and Scaling Teams

A single AI-introduced architectural bug in production costs over $2,500 in senior engineering refactoring hours. RepoGuard protects your codebase for less than a lunch.

Open Source

For public repositories and solo open-source builders

$0/forever

✓ Unlimited Local CLI runs
✓ Automatic .cursorrules generation
✓ CLAUDE.md & Windsurf generation
✓ Architectural Health Score (A+ to F)
✓ Pre-commit drift blocking hook

Install via GitHub

Developer Pro

For independent engineers, contractors & freelancers

$12/month

✓ Everything in Free
✓ Unlimited Private Repositories
✓ Automated PR Review Bot on PRs
✓ Custom Architecture Rule Builder
✓ Automated Secret Leak Detection
✓ Dynamic Live Shields.io Health Badges

Subscribe for $12/mo 🇧🇷 Pagar no PIX (R$ 67 à vista)

⚡ Instant activation • Cancel in 1 click

Engineering Team

For startups & scaling engineering departments

$39/month

✓ Up to 5 Developers included
✓ GitHub Organization-wide enforcement
✓ Strict PR Merge Blocker on CI/CD
✓ SOC2 / ISO Architecture Audit Logs
✓ Slack & Discord alert webhooks
✓ Priority Discord/Email Support

Upgrade Team ($39/mo)

💼 Expense-ready PDF invoice included

🛡️ 14-Day Money-Back Guarantee

⚡ 1-Click Cancellation via Stripe

🧾 Automatic Tax-Deductible Invoices

🔒 100% Code Privacy: Code never leaves your CI

💼 Expensing this to your company? (1-Click Pitch)

"Hi Manager, I'd like to expense RepoGuard ($12/mo solo or $39/mo for team). It audits our PRs to prevent AI-generated architectural rot (bypassing ORMs, duplicate helpers, ': any' types), saving roughly 5-10 hours of senior code review every week."

Frequently Asked Questions

Can I expense this to my company?

Yes, absolutely! Stripe automatically generates itemized, tax-deductible PDF invoices with your company name, VAT/EIN, and address. Almost all software engineering companies have an annual developer tooling budget.

Does RepoGuard send my proprietary code to external servers?

Never. RepoGuard is built with privacy-first engineering. It runs 100% locally on your machine or inside your own GitHub Actions runner. Zero lines of proprietary source code are ever uploaded or stored externally.

Can I cancel anytime?

Yes. You can cancel your subscription at any time with a single click inside your Stripe customer portal. No phone calls, no dark patterns, no cancellation fees.

Stop AI from turning your repo into architectural spaghetti.