DEV Community

Aqeel Abbas
Aqeel Abbas

Posted on

I Built a Full-Stack Freelancer App Using AI — No Manual Coding

The Problem

As a freelancer on Fiverr and Upwork, I constantly
juggled 5+ different tools — spreadsheets for clients,
separate apps for invoices, copy-pasting proposals
manually every time. It was exhausting.

So I built FreelanceOS — a single command center
for everything freelance-related.

What I Built

A full-stack web app with 6 complete features:

🤖 AI Proposal Writer

Powered by Ollama's gpt-oss:120b cloud model.
You fill in your freelancer profile once, paste
a job description, and get a compelling, personalized
proposal following Upwork's official guidelines —
hook, problem statement, solution, proof, CTA.
Under 150 words. Never starts with "I" or "Hello."

💳 Stripe Invoice Payments

Clients receive an invoice link and pay with one
click via Stripe Checkout. Webhook automatically
marks the invoice as Paid. Full payment lifecycle
in one flow.

📧 Professional Invoice Emails

Beautiful HTML invoice emails sent via Resend API
from a verified domain (konvertio.app) with an
embedded "Pay Now" button linking to Stripe checkout.

📊 Revenue Analytics Dashboard

Live charts using Recharts:

  • Monthly revenue bar chart (last 6 months)
  • Revenue by platform donut (Fiverr/Upwork/Direct)
  • Project status breakdown
  • Hours tracked this week

⏱️ Time Tracker

Per-project start/stop timer, weekly summary,
full log history with notes.

👥 Client CRM + Kanban Pipeline

Full CRUD client management with drag-and-drop
Kanban board across 5 stages: Lead → Active →
In Review → Completed → Paid.

How I Built It

Entirely through conversation with MeDo
no manual coding. Here's my prompt strategy:

  1. One big foundation prompt — described the full app structure, UI style, and all 6 pages
  2. Feature-by-feature iteration — each feature got its own dedicated follow-up prompt
  3. Database migration prompt — moved everything from localStorage to real Supabase backend tables
  4. Bug fix prompts — targeted one-liner fixes for specific issues
  5. Polish prompt — animations, empty states, onboarding modal, landing screen

The key insight: treat each MeDo prompt like a
precise engineering spec.
The more specific you
are about data flow, API endpoints, and UI behavior,
the better the output.

Tech Stack

  • Frontend: React + Tailwind CSS
  • Backend: Supabase (via MeDo)
  • Database: PostgreSQL
  • Payments: Stripe
  • AI: Ollama Cloud API (gpt-oss:120b)
  • Email: Resend API
  • Charts: Recharts
  • Platform: MeDo

The Result

A production-quality app that would normally take
2-3 weeks to build manually — done in 48 hours
through conversation alone.

This is what AI-assisted development actually
looks like in 2026.

Try It Live

🔗 https://app-bn4yr7ij1ce9.appmedo.com

Built for the MeDo Hackathon
— If you're a freelancer, give it a try and let
me know what you think!

BuiltWithMeDo

Top comments (0)