How I directed 3 AI Agents to build a Kakeibo App (No-Code Journey)

Published: (December 17, 2025 at 10:49 PM EST)
2 min read
Source: Dev.to

Source: Dev.to

Introduction

My name is Miki, and I’m a developer from Japan. I’m currently learning English, so please forgive any mistakes. I don’t write code manually and I don’t have deep knowledge of Python or JavaScript. Still, I want to share how I built a fully functional Kakeibo (budgeting) app by acting as a “manager” for three different AI agents.

Workflow Overview

Instead of writing code line by line, I managed a workflow using multiple AI tools:

  • Kiro – Requirements and design (Product Manager)
  • Codex – Code generation (Junior Engineer)
  • Google Antigravity – Bug fixing and backend integration (Senior Engineer)

I didn’t code; I coordinated. I didn’t start by guessing; I used Kiro as my Product Manager.

Using Kiro (Product Manager)

Kakeibo is a Japanese method of mindful spending, so the challenge was translating that cultural idea into technical requirements. I told Kiro:

“I want a Kakeibo app that tracks needs vs. wants.”

Kiro created clear user stories and a detailed task list for the app.

Using Codex (Junior Engineer)

I took the task list from Kiro and gave it to Codex. Codex generated the code automatically. Based on the project structure, it used React, Next.js, and TypeScript.

My role was simply to:

  1. Copy tasks from Kiro
  2. Paste them into Codex
  3. Organize the generated files

I acted as the bridge between planning and implementation.

Using Google Antigravity (Senior Engineer)

Connecting the backend was the hardest part for me. I wanted to use Firebase with Google Authentication, but I kept running into configuration errors and SDK issues. This is where Google Antigravity helped. Because it has agentic capabilities, it didn’t just give advice—it:

  • Planned the backend implementation
  • Explained how the SDKs work together
  • Fixed configuration and authentication errors step by step

Resulting Features

After letting the agents do their work, I ended up with a working application that includes:

  • Google login
  • Budget tracking based on the Kakeibo method
  • A simple and clear dashboard

Conclusion

I believe the future of coding is not writing syntax, but orchestrating intelligent agents. I’m still learning English, and I’m still exploring how to build products with AI.

If you have any advice about my workflow—or my English—I’d love to hear it.

Thanks for reading!

Back to Blog

Related posts

Read more »

Kibana – DevSecOps Tool Overview

Overview of the Tool Kibana is an intuitive web‑based data visualization and analytics interface that works on top of Elasticsearch data. It lets users search,...

Amazon Athena in the AWS Periodic Table

!Cover image for Amazon Athena in the AWS Periodic Tablehttps://media2.dev.to/dynamic/image/width=1000,height=420,fit=cover,gravity=auto,format=auto/https%3A%2F...