About this demo

This is a demo site for agentprose, an open-source script that lets AI agents read a blog through WebMCP tools instead of scraping HTML.

Where the articles come from

Every post here is a real issue of Agentic Engineering in Action, the LinkedIn newsletter by Maxim Salnikov. The text is republished unchanged, with the original publication date and a link to the original article at the top and bottom of each post. One issue is a guest post by Einar Ingebrigtsen and is credited to him. To comment or subscribe, use the originals on LinkedIn.

What agentprose adds

The whole integration is one line in the page head:

<script src="/agentprose.min.js" defer></script>

On this site the line also carries a few data- attributes: the site’s sub-path, the prebuilt search index, the newsletter link and the inspector. Once the package is on npm the src becomes a jsDelivr URL.

With it, a WebMCP-capable browser agent gets read-only tools such as get_current_article (the article as Markdown), search_posts, list_posts, get_post, get_table_of_contents, get_related_posts, list_taxonomies and get_site_info. The search index (agentprose.json) is built in CI by agentprose build after hugo.

Click Try the agent tools in the header to open the inspector and call the tools yourself. In Chrome with chrome://flags/#enable-webmcp-testing turned on, the inspector uses the browser’s own document.modelContext.

Not for search engines

This copy exists only as a demo. robots.txt disallows all crawlers and every page carries <meta name="robots" content="noindex, nofollow">, so the originals on LinkedIn stay the canonical versions.