Skip to main content
wondeya turns a landing page into a conversation. A visitor asks a question and an AI agent answers by rendering interactive visual components (a gallery, a timeline, a set of headline figures, a contact form) instead of paragraphs of chat text. Every answer is built from the material the business published: the documents you upload, the pages you sync, the short facts you write by hand, and the images you register. A static landing page shows every visitor the same thing and hopes it covers their question. A wondeya page composes the page each visitor needs, one question at a time.

​
What a visitor sees

A published wondeya page is a real web page, server-rendered, with your colours and your typeface. It carries a visible notice that the answers are generated by an AI, and an input where the visitor types. When they ask something:
  1. Sections appear as placeholders the instant the agent decides what the answer looks like.
  2. Each placeholder is replaced in place as its component finishes composing: an image with a caption, a set of steps, an expandable FAQ, a form.
  3. A one-line written reply arrives alongside the components.
Follow-up questions refine what is on screen instead of restarting the page.

​
What you configure

You configure three things. An agent is the brain: the knowledge, the images and the identity it answers with, configured once and reusable. A project is one site: its your-site.wondeya.app address, its domains and its brand, including the theme. A page joins them: an address inside a project plus the agent that answers there.

Knowledge

Documents, URLs and curated claims, owned by the agent. The only material an answer may be built from.

Assets

The images the agent may show, each described so the agent knows when it fits.

Theme

Five colours, two typefaces and one corner radius, from a set the server owns. Part of the project’s brand.

Page settings

The page’s address in the site, the languages, who the agent speaks for, and the act of publishing.

​
Two ways to run it

  • A hosted page. Each project is a site with its own address at your-site.wondeya.app, proposed from the business name and correctable; publishing its home page opens it to the world. A workspace can run several sites.
  • A widget on your own site. One script tag puts the same agent on a site you already have, inside an isolated iframe on a separate origin. See the widget.
Both talk to the same engine, with the same knowledge and the same rules.

​
Two audiences, one product

Teams without engineers

Do the whole job in the web console: create the site and its agent, upload knowledge, upload images, pick colours, publish. No code at any step.

Developers

Drive the same operations over the REST API, or as MCP tools from Claude and other AI clients, authenticated with OAuth. Or embed the widget in your own front end with one script tag.

​
What makes the answers safe

The model never draws the page. It selects: a component type from a closed catalog, an asset by id, a link by id. The server then resolves every id into real data after validating it. A URL is not even expressible in the model’s output, which is what makes it safe to render an AI answer on a public page in a customer’s name. How it works walks through the pipeline; Security and privacy covers isolation, the AI notice and deletion.

​
Vocabulary

​
Where to go next