AI Content Assistant MVP — From Dev Setup to Live Service in One Day

Date: 2026-04-24

The AI Content Assistant MVP for small business owners is live. Built in one day.

Dev Environment

Set up VS Code and Claude Code in the morning. Kept the project completely separate from the stock auto-trading project to avoid conflicts. Isolated to C:\projects\ai-content-mvp.

Created the project with Next.js 16 + TypeScript + Tailwind CSS. First push to a GitHub private repo done.

Deployment Pipeline

Connected Cloudflare Pages for auto-deployment. One git push triggers the full build and deploy pipeline. The key was solving the deployment issue with static export mode.

Connected the aicontent.chulbuji.com subdomain via Cloudflare DNS CNAME. SSL applied automatically.

Screen Implementation — Claude Design Handoff

Used Claude Design Handoff to pass the prototype directly into Claude Code. 8 screens were implemented automatically: onboarding, input, loading, results, editing, and templates.

API Integration

Connected OpenAI GPT-4o API via a Cloudflare Worker. Structured prompts by business type and content goal.

Business Type	Content Goal
Café / Hair Salon / Craft Studio / Retail	New product / Grand opening / Event / Return visit

Confirmed working generation of: SNS copy, blog posts, short-form scripts, hashtags.

Deployment Automation

GitHub Actions pipeline deploys both Pages and Worker simultaneously. No manual wrangler commands — one git push handles everything.

Today’s Numbers

• Total dev time: approx. 1 day
• Screens implemented: 8
• Time from git push to live: 23 seconds
• Current infrastructure cost: ₩0

Next

Real-user test with 5 small business owners. A feature no one uses isn’t finished.