Stop vibe-coding chaos.
Enforce real architecture & security.
Automatically.
Kozo is the "Senior AI Architect" that lives in your project. It enforces the laws of clean code, security, and context—automatically—across Cursor, Lovable, Emergent, and every other agent you use.
Watch your AI Agent say "NO" to Technical Debt.
"I asked Emergent to add a function to a 1,200-line file. Because of the Vibe Constitution, it refused. It calculated the violation, explained why it was bad for my project, and offered 3 modular alternatives instead. This is governance in action."
If any of this sounds familiar,
you're in the right place.
- •You get a working app super fast… then adding one tiny feature breaks three other things
- •Files keep growing — 1,000, 2,000, 10,000 lines — and you don't realize it's a problem until everything slows down or the agent tells you
- •You paste an API key "just for now"… and later panic when you remember it's still in the code
- •The agent forgets what you decided three chats ago and starts contradicting itself
- •You keep fixing the same kinds of issues over and over… and wonder why vibe coding feels slower than hand-coding sometimes
Most beginners don't see these traps coming. We didn't either — until we did.
That's why Kozo exists: Simple rules that teach your AI (and you) how to build clean, safe, scalable apps from the start.
"Once the rules are in place, the agent literally refuses to make things worse."
How It Works
Super simple 4-step flow for beginners
Answer 5 easy questions
What are you building? What tools do you use? What's annoying you most? Takes ~60 seconds.
Get your beginner-friendly constitution
We create simple, clear rules tailored to you: keep files small, no secret keys in code, basic security, memory helper, and "rescue mode" if your project is already a mess.
Drop the files in your project
One zip file. Copy-paste into Cursor, Lovable, Emergent, Bolt.new — done in under 2 minutes.
Watch your AI get smarter (and nicer)
Agents start refusing bad ideas. They suggest proper splits instead of bloating files. They remind you about security.
Before — agent nagging about refactor
After — agent suggests modular approach
Meet the Sentinel:
Your 24/7 Architectural Watchdog.
Rules are only useful if they're followed. Sentinel connects to your GitHub repo to ensure your AI (and your team) never drifts from the Constitution.
ARCHITECTURAL ALERT: This PR introduces a 650-line file in /services. This violates the 500-Line Law. Please refactor before merging.
Automated PR Blocking
Sentinel acts as a "Quality Gate" in your CI/CD pipeline. It automatically flags (or blocks) Pull Requests that violate your core laws.
Drift Detection
AI agents can be "pushy." Sentinel detects when an agent tries to ignore your .mdc files and alerts you before the technical debt becomes permanent.
Shadow-Secret Protection
Continuous scanning for API keys and PII that might have been "hallucinated" into your codebase by an AI during a late-night session.
The Observatory (Dashboard)
A high-level view of your "Project Health Score." Watch your technical debt shrink in real-time as your Constitution takes hold.
Beta testers get lifetime Pro — so you'll see the full power first.
Your code is safe with us.
We know trust matters when uploading code.
That's why we built it this way from day one:
Local Scan
Our scanner runs 100% in your browser. We never see your source code; we only analyze the "Vitals" (file size, secrets, tech stack). No storage, no sharing, no training our models on your code. Guaranteed.
You own everything. Always.
Your constitution, your rules, your project. We just help you organize it.
Ready to stop guessing and start building better?
We're opening 20 beta spots for coders who:
- ✓Are actively vibe coding (even if it's messy)
- ✓Use Cursor, Emergent, Bolt.new or similar
- ✓Want cleaner code without slowing down
- ✓Can share quick feedback once a week
Spots are limited — apply today.
Privacy & Security FAQ
Is my source code uploaded to your servers?
No. We use a "Local-First" architecture. When you select a project, your browser performs a local scan using our open-source scanner. The raw source code is read into temporary memory, analyzed for architectural patterns, and then immediately discarded. Only the "Vitals" (metadata like file sizes and line counts) are ever sent to our server.
What exactly are "Vitals"?
Vitals are a high-level summary of your project's health. They include: Structural Metadata (file names, folder depths, language breakdown), Complexity Metrics (line counts and "Spaghetti" hotspots), and Security Alerts (notifications that a secret was found, but never the secret itself).
How does the Secret Scanner work if it doesn't upload code?
The scanner looks for patterns (Regex) in your browser. If it finds something that looks like an API key or a password, it records the type and the location. Before that information is sent to us, we run a redaction script that scrubs the value entirely. Example: Instead of seeing your actual key, our server sees: sk-[REDACTED].
Can I use this for enterprise or highly sensitive projects?
Yes. Because we never ingest your source code, Kozo is compatible with most corporate security policies that forbid "Code-Leaking" into third-party AI tools.