your AI CTO
Tier 1 kits are live: Next.js + Supabase

Your AI writes the code. 
Now it has a CTO. 

AI CTO decides what to build first, then sets up a production-ready project with the structure, rules and launch checks already in place. The agent you already use builds on that, not on an empty folder.

See how it works

Works with Claude Code, Cursor and Codex CLI.

Every Tier 1 project starts on

  • Next.js 16
  • TypeScript
  • Supabase
  • Postgres + row level security
  • Tailwind v4
  • shadcn/ui
  • TanStack Query
  • Zod
  • Vercel

01 / 05

Say what you want to build.

Plain English, the way you'd tell a co-founder. No spec, no jargon.

New conversationAI CTO

What are we building?

Describe it the way you'd tell a co-founder.

ThinkResearchSearch

02 / 05

It thinks before it answers.

It reasons through the idea, looks up what it doesn't know, and asks the one question that changes the build.

Dabba: office tiffin subscriptionsAI CTO
I want to run a tiffin subscription for office workers in Pune. They pick meals for the week and pay monthly.
Thinking…
Who pays decides whether there are one or two kinds of user.
Reply to your CTO…
ThinkResearchSearch

03 / 05

Then it cuts the scope.

What to build now, what waits, and what stays out of version one, on a stack it can actually install.

Dabba: office tiffin subscriptionsAI CTO
I want to run a tiffin subscription for office workers in Pune. They pick meals for the week and pay monthly.

Employees pay. One office to start, about forty people.
Writing your plan…
Reply to your CTO…
ThinkResearchSearch

04 / 05

One command. A real project.

doctor checks your machine. Then the kit lands: the structure, the rules, the guides and your plan.

Terminal~/projects
Waiting for the scaffold…
05 / 05
Your agent builds inside the rules.
Claude Code, Cursor or Codex reads AGENTS.md first. When it slips, the checks catch it before you do.
Claude Code~/projects/dabba
›Build the weekly menu from plan.md
Working…
running in dabba/
What lands in your folder.
Not a template to fill in. A working app, with the rules written down where your agent reads them, and your product built on top.
  1. 01The rules
    AGENTS.md, the conventions and the lint rules. Your agent reads them before it writes a line.
    AGENTS.md
  2. 02Guides and checklists
    Ten step-by-step module guides, and three gates: before you build, before you launch, security.
    docs/checklists/pre-launch.md
  3. 03A working app
    Sign-in, the API, three database clients and 61 components, already wired and building.
    src/lib/db/supabase/server.ts
  4. 04Supabase and Vercel
    Postgres with row level security, file storage and deploys, set up step by step in INITIALIZE.md.
    supabase/migrations/example_rls.sql
  5. 05Your product
    What your agent builds on top of all that: Dabba's menus, subscriptions and the kitchen's count.
    src/app/(app)/menu/page.tsxby your agent
Same AI. Different foundation.
The same six mistakes, sent from your agent towards your users.
Blank folder: agent → deploy → users. All six reach your users.
With the AI CTO kit
  1. Your agent
    Claude Code · Cursor · Codex
  2. AGENTS.md rules
    While it writes
    •  A table with no access rules
    •  A list that loads every row
  3. npm run check
    Before it's done
    •  Code in the wrong place
  4. Pre-launch gate
    Before real users
    •  Sample data left on a screen
    •  No spend limit
  5. Security gate
    Before launch
    •  The admin key in the browser
  6. Deploy
    Vercel
  7. Your users
    0 incidents · 1 change live
Your agent is ready.
Give it a CTO.
Describe your idea tonight. Read the plan, install the project, and let your agent build on a foundation that says no.
Not ready to sign up? doctor checks your machine in seconds, no account needed.