Using Laabam.site from ChatGPT or Claude

Laabam.site speaks MCP — the Model Context Protocol. It is the standard way an AI assistant connects to an outside service. Connect your Laabam.site account once, and ChatGPT or Claude can answer questions about your websites, and build new ones for you.

Two things make this different from asking an assistant to "make me a website":

• It can see your real account. How many enquiries came in this week, whether your site is live, what your plan is. Real numbers from your dashboard, not guesses.
• The assistant writes the code, and you pay nothing to Laabam.site for it. When ChatGPT designs a site and sends it here, no Laabam.site AI agent runs, so no Laabam.site credits are spent. You are already paying OpenAI or Anthropic for that model; you do not pay us twice.

The two ways to build a website

There are two paths and they cost different things. Pick on purpose.

Neither is better. The dashboard is faster if you want a finished thing. The assistant is better if you want to argue with it about the hero image for twenty minutes.

What the assistant can see

Once connected with read access, an assistant can look at:

• Your websites, whether they are live, and your custom domains
• How many enquiries arrived, how many are unread, and the messages
• Visits, open SEO work, your credit balance and your plan

What it can never see

• Phone numbers left by your visitors. These are removed before anything leaves Laabam.site. The assistant is told how many enquiries included a number, never the number itself. You read those in your dashboard.
• Your uploaded images
• Your password

What needs your permission

Two things are separate, unticked boxes on the screen where you connect. Asking is not the same as being granted — you decide each one.

• Propose new websites. Lets the assistant send finished code to your account. A proposal creates nothing and spends nothing; it waits in your dashboard for you.
• Read your website's code. Lets the assistant read the pages you already have, so it can match your existing design instead of guessing at it. Your markup is sent to whoever runs that assistant. This one needs a paid plan.

Nothing gets published without you

This is the rule the whole thing is built around.

When an assistant sends a website, it does not go live. It records a proposal, and the proposal waits in Settings → AI connections until you open it, read the code, and press Publish it.

That click is the publish. There is no second step after it — the site goes to a public address immediately — which is exactly why you should read the code first. The proposal card has a Review code button that shows you every file and a preview of the page.

The reason this matters is not that the assistant makes mistakes. It is that assistants read things: web pages, PDFs, emails. Any of those can contain text written to manipulate the assistant rather than to inform you. Your click is the thing that cannot be faked by a document the assistant happened to read.

Getting set up

• Connect ChatGPT step by step — the developer-mode connector, with the two settings people get stuck on
• What each tool does — the full list, and which need a paid plan

Cost, plainly

The only thing that spends Laabam.site credits is Laabam.site's own AI doing work.