WebMCP AI Visibility is a free WordPress plugin that describes a site’s public content in a structured way for AI agents. It relies on WebMCP, a way of describing a site’s content and functions as tools for compatible AI agents and browser-based clients. It’s built by BeeClear – an agency focused on search engine and LLM-answer visibility and WordPress website development. What sets the plugin apart is precise control over content visibility. You decide which content types and tool groups an agent gets to see. Here’s how that mechanism works and how to set it up in the dashboard.
Table of contents
- Why decide what an AI agent sees on your site?
- How does WebMCP AI Visibility control content visibility for AI agents?
- Which content types should you expose to AI agents?
- What does one content-type list change across the whole plugin?
- When should you turn off a given tool group for agents?
- Will an AI agent see private or password-protected posts?
- Does the llms.txt file use the same visibility settings?
- How do you check what an AI agent actually sees?
- How do you set content visibility on your own site?
Why decide what an AI agent sees on your site?
It’s worth deciding what content is visible to an AI agent, because not every piece of content in WordPress is meant for outside audiences. A site accumulates helper pages, technical subpages, and content types added by plugins. Some of them are thin content or support internal processes. Mixed in with your articles and offer pages, they can blur an agent’s picture of the company.
Even read-only tools matter here. They don’t let anyone submit forms or perform administrative actions. But they do make it easier to systematically browse public content, post by post. That’s why you set the scope of exposed data yourself, from the dashboard.
How does WebMCP AI Visibility control content visibility for AI agents?
WebMCP AI Visibility controls content visibility for AI agents on three levels, all set from the WordPress dashboard. The first is the list of content types exposed to agents. The second is the tool groups you turn on or off with a single field. The third decides whether read-only tools are public or reserved for the administrator.
The entry point for an agent is the REST manifest, available at the /wp-json/beeclear-webmcp/v1/manifest path. The plugin only lists tools from active groups there. An inactive group disappears from the site’s description. A request for its tool then fails with an unknown-tool error.
The third level is especially useful at the start. Open access to read tools is enabled by default, but you can turn it off temporarily. Only the administrator can use them then. That lets you review the plugin’s responses calmly before you expose them to outside agents.
Which content types should you expose to AI agents?
Expose the content types to AI agents that best describe your offer and expertise. By default, the plugin exposes posts and pages. In the settings list, though, you’ll see every public content type on the site, including ones added by other plugins. You only check the ones that should feed the search and content-retrieval tools. An agent asking which content types are available will also get only your selection back.
The plugin itself hints at what to avoid. It’s not worth exposing types with thin, internal content, or pages that support private processes. The same rule applies to objects that were never meant to be discovered automatically. Articles, guides, and documentation are usually a good choice, while technical helper types generally are not.
What does one content-type list change across the whole plugin?
A single content-type list drives every channel the plugin uses to serve content to agents. You don’t set search, Markdown, and taxonomies separately. The type selection works consistently across all of these:
- post search only returns content from the checked types,
- fetching a single post returns no result for a type outside the list,
- the Markdown version for clients sending an Accept: text/markdown header covers only exposed types,
- taxonomies and terms only appear for an exposed content type,
- media attached to a post from a hidden type doesn’t show up in media search results.
Handling requests outside the list matters just as much. An agent may ask for a content type that isn’t checked. Search then returns an empty response. The plugin doesn’t fall back to querying every available type instead. That way, a hidden type stays hidden, no matter how the question is phrased.
When should you turn off a given tool group for agents?
Turn off a given tool group for agents when that area of the site shouldn’t be visible to them. The plugin splits tools into five groups, each controlled by its own field in the settings. All of them are active by default, so review them before opening the site up to agents. Here’s what each one exposes:
| Tool group | What the agent sees | When to enable it |
|---|---|---|
| Content | search and retrieval of published posts, pages, and selected content types | when reading content is part of the agent’s job |
| Navigation and taxonomies | menu locations, menus, taxonomies, and terms | when the agent needs to understand the site’s architecture |
| Media | attachment metadata, image URLs, MIME types, and alt text | when the agent needs to analyze images or downloadable files |
| Contact Form 7 | field schemas and the purpose of forms | when forms are part of the conversion path |
| Custom forms | forms described via CSS selectors | only after you’ve verified the selectors on the live site |
Contact Form 7 forms have an extra, more granular switch. On the forms screen, you can turn off visibility for a single form while leaving the rest unchanged. That’s useful when only some of your forms make sense for an agent.
You can make a similar selection for custom functions, meaning selected REST endpoints on the same site. Mark a diagnostic function as a developer tool, and the plugin adds a debugging annotation to it, supported since Chrome 156. General-purpose agents can then filter such tools out, so test endpoints don’t get mixed in with the rest.
Will an AI agent see private or password-protected posts?
An AI agent will not see private or password-protected posts through the plugin. Content tools and the Markdown version only cover published content from exposed types. A password-protected post is excluded, even if it belongs to a checked type. Media protection works the same way, since an attachment linked to a password-protected post doesn’t show up in results.
A separate boundary applies to tools that change the state of the site. They require a standard WordPress REST nonce and a logged-in user context. Without logging in, only read tools work, and only if public access is left enabled. Open endpoints also carry a simple rate limit, which curbs accidental, repeated load.
Does the llms.txt file use the same visibility settings?
The llms.txt file doesn’t use the same visibility settings, because WebMCP AI Visibility gives it its own, separate content-type list. That list only applies when the plugin itself generates llms.txt and llms-full.txt. By default, only pages go in there – lasting, structural content, not every blog post. So you can expose your blog to search tools while keeping llms.txt limited to your offer subpages. When another plugin handles llms.txt, this setting has no effect on it.
How do you check what an AI agent actually sees?
You can check what an AI agent actually sees on the plugin’s test screen. Quick tests load the manifest, discovery files, and the most common read tools. The result is shown as raw JSON, exactly as the agent receives it. After changing content types or tool groups, you immediately see what disappeared from the response. That turns your settings from a declaration into a verified fact.
The tests also cover the llms.txt, agent-manifest.txt, and agents.txt files, as well as the OpenAPI schema. You can check all of these documents in one place, with no external tools. Only once the responses match your intended plan do you open the configuration up to agents.
How do you set content visibility on your own site?
You can set content visibility on your own site within a few minutes of installing the plugin. Download WebMCP AI Visibility from the WordPress.org directory and activate it in the dashboard. Then review the content-type list, turn off the tool groups you don’t need, and save your settings. Finally, run the tests and compare the results against your intended scope.
The plugin doesn’t change the look of your site or your theme templates. Visitors see the site as before, and agents see only what you point them to. You can change the settings at any time, and the next test will immediately show the effect. You’ll find the full feature description on the WebMCP AI Visibility plugin page.
