The Vibe Coding Stack I Actually Use

Everyone has an opinion on the best tools for building with AI. Most of those opinions come from people who haven't actually shipped anything. I have - so here's the exact stack I use to go from idea to live app, what each tool does, why I chose it and what it costs.

Step 1: Start with an LLM to flesh out the idea

Before I open any building tool, I have a conversation with an LLM - usually Gemini or ChatGPT, though honestly it could be any of them. The model doesn't matter as much as the process.

I describe what I want to build. What the app does, who it's for, what features it needs. I treat it like briefing a collaborator - the more specific I am, the better the output.

But here's the part most people skip: once the idea is solid, I ask the LLM to generate the prompt I'll use in the next step. That handoff prompt becomes my blueprint. It means I'm walking into the design phase with a clear, structured brief rather than a vague idea.

Step 2: Design in Google AI Studio

Google AI Studio is where the idea becomes something visual. I use it to design the UI, map out the pages and get a feel for how the app will actually work before I build a single line of code.

Why Google AI Studio specifically? Right now it is extraordinarily generous with its free usage. Compared to tools like Lovable or Bolt - where you can burn through credits quickly and hit a paywall - Google AI Studio lets you iterate freely without watching a credits counter tick down.

That may change. But right now, for the design phase, it's hard to beat.

I stay in Google AI Studio until I have enough pages and enough of the UI that I feel confident handing it off. That's the moment I move to the build phase.

Step 3: Build in Google Antigravity using Claude Code

Google Antigravity is an IDE (Integrated Development Environment - think of it as your coding workspace) but not in the traditional sense. It's an agent-first platform, which means instead of writing code yourself, you're acting as the architect directing AI agents to build for you. You manage the vision and the outcomes. The agents handle the construction.

It's completely free during public preview, which makes it an easy choice.

Inside Antigravity, I use Claude Code to bring the app to life. Claude Code is Anthropic's coding agent and in my experience it is currently the strongest option in the market for this kind of work. It thinks carefully, follows complex instructions well and produces clean, functional output.

You'll need at least a Claude Pro subscription to access Claude Code, which is $20/month at the time of writing. Worth every penny.

What does this actually cost?

Here's the honest breakdown:

• Google AI Studio: free (generous usage limits, paid plans available if you need more)

• Google Antigravity: free during public preview

• Claude Pro (required for Claude Code): $20/month

• Lovable / Bolt / Replit: optional alternatives, all have free tiers but paid plans start from around $20/month

The total minimum cost to run this stack is $20/month.

Compare that to hiring a developer - which realistically starts at several thousand pounds for even a basic app - and the shift is pretty extraordinary.

Pricing is correct at the time of writing and subject to change.

Why this stack works

Each tool earns its place. Google AI Studio gives you a generous, visual design environment to shape the idea. Antigravity gives you a proper agent-first build environment. Claude Code gives you the most capable coding agent available right now.

And the workflow itself - using an LLM to generate your handoff prompt before you ever open a design tool - is the bit most people aren't doing. It means every step has a clear brief and you're not figuring things out as you go.

That's the stack. Just the tools I actually use and why they work together.

Want to learn how to build your first AI app using this exact workflow? Come and join us at Vibe Coding Lab. It's free and we build things together.