Go from a single prompt to a deployed Wix-managed Astro project. You describe what you want to build, and a coding agent on your computer creates the business, scaffolds the Astro project, and publishes the site. You continue from the project folder with the Wix CLI.
This is the Astro path, with the Wix CLI.
With a single prompt, your agent can build and deploy a complete, working site backed by Wix business solutions: a store, a booking system, a blog, and more.
Describe what you want in plain language:
The agent reads your prompt, picks the solutions your site needs, and installs them for you. This is a starting point, not a limit: because the skill reads Wix's live documentation, your agent can discover and wire up any Wix business solution, not just the common ones. If you need something specific, name it in your prompt, or paste a link to the relevant Wix documentation for the agent to read. Learn more about featured business solutions for Wix Headless.
Tip: The more detail you give, such as the desired pages, data, and design, the more complete the result. For ready to adapt examples, see the prompt recipes.
Make sure you have:
The Wix Headless skill tells your coding agent how to create the business, scaffold an Astro project, and publish it on Wix.
Caution: This path always provisions a new business and site. If you already have a project, from the self-managed quick start or a dashboard-created headless client, this flow can't attach to it. If you paste an existing siteId or clientId into your prompt, the agent won't use it, and you'll end up with 2 separate, unrelated Wix projects. To use a project you already have, leave the skill URL out of your prompt. Call the JavaScript SDK or the REST API with that project's site ID and client ID.
In your coding agent, describe what you want to build and include the skill URL. For example:
The agent provisions the business and site, scaffolds the Astro project in a folder on your computer, and publishes the frontend. For more detailed, ready-to-adapt prompts, see Prompt recipes. To understand what happens between your prompt and the deployed site, see How it works.
You can tell the skill is working when the agent asks you to log in to Wix, reports that it provisioned a site, and names the business solutions it installed.
Your project appears in your Wix dashboard alongside your sites. Open your dashboard to view the live URL, manage the site, or connect a custom domain.
The agent saves the Astro project in a folder on your computer. Ask it where it saved the project. You continue from that folder with the Wix CLI. Edit Frontend doesn't open these Astro sites in another tool.
The single-prompt skill is enough to scaffold and deploy. For ongoing development, install the Wix plugin in your AI client. It adds Wix Skills and the Wix MCP together, giving your agent persistent Wix context so it can search Wix docs, scaffold code, and call Wix APIs directly from your chat.
Keep prompting the agent in the project folder to add features and functionality. Define your data model, add authentication, build UI, and redeploy.
To preview and publish from that folder, run npx wix preview, then npx wix release.
You don't need to stay in the same chat to keep working. Start a new chat in the project folder and tell the agent what you want to change. Site Settings in the dashboard has Publish and Unpublish. After you release from the project folder, publish further updates with npx wix release.
A lot happens between your prompt and the deployed result. You don't run any of it yourself, but knowing the sequence helps you follow along, tell when something has gone wrong, and steer the agent toward the site you want.
The skill runs with little back-and-forth, but the agent pauses when a choice is genuinely yours to make. Common examples:
Answer in plain language. If the agent doesn't ask about something you care about, such as a specific business solution, a data schema, or a design direction, say so in your prompt up front instead of waiting for the agent to ask.
For the best result, follow these best practices:
Because the agent infers a lot from a short prompt, the first result doesn't always match what you pictured. These are some common gaps and how to close them:
A short prompt gets you started, but a detailed one gets you a finished site. The recipes below show how to structure a thorough prompt: group your requirements into functionality, pages, and design, name your data up front, and ask the agent to seed realistic example content.
Last updated: 12 November 2026