The vibe coding toolkit: Choosing the right platform for your next build

In our video series, **Is everyone a developer now?**, we’ve been exploring the shift toward building apps using natural language. It is an exciting time to be in the ecosystem, but a common question keeps popping up: with the explosion of new AI-assisted development platforms, which one should you actually pick to build a Xero-integrated app?

I recently sat down with Regan to map out the current landscape. We wanted to move past the hype and look at where these AI coding tools actually sit in your workflow. Rather than a deep dive into every tool, we’re sharing recommendations based on your proficiency level and our own experience with the platforms.

[embed]

Categorising the vibe coding landscape

During our chat, Regan helped me group these tools into two distinct categories based on how much of the technical heavy lifting they handle for you.

1. The all-in-one platforms

Platforms like Lovable and Replit are “everything-in-a-box” experiences. They do more than just help you write code; they manage the entire stack on your behalf.

When you use these, you aren’t worrying about setting up servers, managing databases, or configuring CI/CD pipelines. The platform makes those technical decisions for you. This is a great starting point if you want to move fast and do not want to get bogged down in infrastructure.

2. The high-control power tools

Then you have tools like Cursor and Claude Code. These offer significantly more power and flexibility, but they also require you to take the training wheels off.

If you use these tools, you are back in the driver’s seat for hosting and architecture. You will need to understand concepts like DNS and caching. These environments keep the code front and centre.

Matching the tool to your proficiency

Regan uses a simple 10-point scale to help people decide where to start. On this scale, 0 is a total beginner and 10 is a senior engineer. Here is how we see the current lineup of apps we’ve used:

• Lovable (Level 1–2): This is the fastest path from a blank screen to a working app. If you are a small business owner or an accountant who needs to solve a specific problem without learning full-stack development, start here.
• Claude Code (Level 4–5): This tool works natively with Anthropic’s models. It requires some comfort with a terminal, but it is incredibly fast for building based on specific technical requirements.
• Cursor (Level 6): Since this is built on Visual Studio Code, you are looking at the source code at all times. It is the gold standard for anyone who wants to learn how to debug while they build.

The trade-off: Cost vs convenience

We also discussed the “hidden” costs of these tools. While an integrated platform like Lovable might have a higher monthly subscription, Regan pointed out that you are paying for the removal of complexity.

With “power tools” like Cursor, your AI usage costs might be lower, but you will eventually have to pay separate bills for your own hosting and database providers. You are essentially trading your time and technical knowledge for a lower price tag.

Recognising the security risks of AI generated code

One thing Regan and I really wanted to drive home is that no matter how much the AI does for you, the responsibility for the data remains with you.

When you build a Xero API integration, you are handling sensitive financial information. We have a shared duty of care for our customers. Whether you are vibe coding a small script or a full-scale app, you must be a responsible custodian of that data. That means keeping your API keys secure and staying on top of security patches.

What is your next build?

The best way to figure out which tool fits your style is to just start building. If you need a hand getting started, we have put together prompts and guides over at the **Xero AI Toolkit**.

Happy building!