Everything Agentimus publishes for AI tools lives in a handful of plain files served straight off your own domain. This page lists every one of them, shows you how to open and read each in a browser, and explains when each is served and how to turn it on or off.

None of these files change what your human visitors see. They are parallel, machine-friendly editions of content you already publish — a clean copy for the assistants and agents that read your site on a person’s behalf.

The files at a glance

File	Address on your site	On by default?	Where to switch it
AI page guide	/llms.txt	Yes	Settings → Discovery → AI page guide
Full-text edition	/llms-full.txt	Yes	Settings → Discovery → Full text for AI
Per-page Markdown	add .md to any page URL	Yes	Settings → Discovery → Plain-text versions
Change feed	/agentimus-changes.json	Yes	Settings → Discovery → Change feed
Discovery document	/.well-known/discovery.json	Always on	(no toggle — served while the plugin is active)
Agent card	/.well-known/agent-card.json	Always on	(no toggle)
MCP / tools manifest	/.well-known/mcp.json	Always on	(no toggle)
XML sitemap (fallback)	/agentimus-sitemap.xml	Only if nothing else provides one	Settings → Discovery → Sitemap (backup)

The three /.well-known/ documents have no on/off switch: they describe your site’s identity and are safe and cheap to serve, so Agentimus always publishes them while it is active. The text files (llms.txt, llms-full.txt, Markdown) each have a toggle on the Discovery tab of the settings screen. The fallback sitemap is special — see its section below.

Only the content types you have chosen appear in these files. By default that’s your posts and pages; other content types (products, custom post types) are opt-in on the Content settings. Draft, private and password-protected content is never included — Agentimus only ever exposes what your normal site already shows to the public.

How to view any of these files

In a browser. Most of these are just web addresses. Type your site’s address followed by the path — for example https://example.com/llms.txt — and the file opens as plain text or JSON you can read directly. The settings screen also has a Live endpoints box with ready-made links to your llms.txt, llms-full.txt and robots.txt, and the Discovery hub lists the /.well-known/ documents.

The one exception — the “Accept header” versions. Some of what Agentimus serves is triggered by a request header rather than a distinct URL. A browser’s address bar can’t send that header, so to see those you use a command-line tool such as curl:

# Ask any page for its Markdown edition using the Accept header
curl -H "Accept: text/markdown" https://example.com/about/

If you’re not comfortable with the command line, you don’t need it — the same Markdown is always available at the plain .md address described below, which any browser can open.

The AI page guide — /llms.txt

/llms.txt is a one-page map of your site written for AI assistants. It’s the first thing many of them look for. Open https://yoursite.com/llms.txt and you’ll see a tidy, human-readable outline:

• Your site’s name and tagline at the top.
• A one-line note telling agents they can fetch any page as Markdown (by adding .md or sending the Accept header).
• A pointer to your discovery document (/.well-known/discovery.json), so an agent arriving here can find your structured machine profile.
• An About block drawn from your Identity settings — who you are, who the site is for, what it is not, and your areas of expertise.
• A section for each content type you publish (Pages, Posts, and any others you’ve enabled), listing each item as a link with a short excerpt.
• A Topics section built from your categories.
• An Optional section linking the full-text edition (if it’s on), your RSS feed, and your sitemap (if one is detected).

When it’s served: on every request to /llms.txt while the AI page guide toggle is on. It’s built once and cached for about an hour, and rebuilt automatically whenever you change your content, so it’s always current without slowing anything down.

Turn it off: Settings → Discovery → untick AI page guide.

The full-text edition — /llms-full.txt

Where /llms.txt is a list of links, /llms-full.txt is the actual text: your profile followed by the complete content of each page and recent post, all concatenated into one document an assistant can read in a single pass. Open https://yoursite.com/llms-full.txt to read it.

Because this file can get large, and because a crawler can ask for it at any moment, Agentimus builds it under a strict size budget so it never strains your server or becomes too big to ingest:

• A total byte budget — about 1 MB by default. Generation stops cleanly at that ceiling.
• A per-item cap, so one enormous page can’t dominate the file (it’s skipped and stays reachable at its own .md address).
• A short time limit on how long the build may run.

When any of these limits is reached, Agentimus stops at a clean page boundary, appends a short note explaining the file was truncated, and points the reader back to the /llms.txt index and the per-page .md addresses for anything left out. The document is always valid — it’s never cut off mid-sentence.

You control how much goes in with Posts in /llms-full.txt on the Discovery tab (default 50; this caps posts and other content types — your pages are always included up to 50). As you change that number the screen shows a live estimate of the resulting file size and warns you if it would exceed the budget. If the estimate is over budget, lower the post count or simply rely on the /llms.txt index instead.

When it’s served: on every request to /llms-full.txt while Full text for AI is on. It’s cached, and after you edit content Agentimus quietly regenerates it in the background so visitors rarely wait for a fresh build.

Turn it off: Settings → Discovery → untick Full text for AI.

Per-page Markdown — the .md address and the Accept twin

For any single page or post you’ve included, Agentimus can hand an assistant a clean, text-only Markdown version — just the words, with none of your theme’s menus, sidebars or widgets. There are two ways to reach the exact same content:

1. Add .md to the page’s address. A page at https://example.com/about/ is also available at https://example.com/about.md. This works in any browser.
2. Send the Accept: text/markdown header to the normal page URL. An agent that prefers Markdown gets the Markdown edition back from the same address a human would get the HTML page from (see the curl example above).

The home page and archives have a Markdown twin too: https://example.com/index.md (or the Accept header on your front page) returns the same site map as /llms.txt.

Each page’s Markdown starts with a small header — the title, and for posts the date, author and categories — followed by the page’s canonical link, a Topics line (if you’ve set topics for that page), and then the body converted faithfully from HTML to Markdown, preserving headings, lists, links, quotes and code.

Agentimus also advertises the current page’s .md twin in the page’s response, so a well-behaved agent can discover it without guessing.

When it’s served: whenever the Plain-text versions toggle is on and the request targets an included page — either via its .md address or the Accept header. Negotiated (Accept-header) responses are never cached, since they share a URL with the normal HTML page.

Turn it off: Settings → Discovery → untick Plain-text versions.

The change feed — /agentimus-changes.json

Assistants shouldn’t have to re-read your whole site to notice one edit. The change feed is a compact JSON list of your recently added, updated and removed pages, so an agent can ask “what’s new since last time?” and fetch only that.

Open https://yoursite.com/agentimus-changes.json to see it. Each item carries the page’s title and URL, when it changed, whether it was created, updated or deleted, a link to its Markdown twin, and a link to its canonical REST resource. An agent that already has a copy of your site can add ?since= with a date or timestamp — for example /agentimus-changes.json?since=2026-07-01 — to receive only what changed after that moment.

Removals are included too: unpublish, trash or delete a page and it appears once as a deleted item (for about 90 days) so an assistant knows to drop it from its own copy — something a plain sitemap can’t tell it.

It is on by default and advertised in your discovery document, so agents find it automatically. Turn it off under Settings → Discovery → Change feed.

The discovery documents — /.well-known/

The /.well-known/ folder is a web-standard location for machine-readable metadata. Agentimus serves its discovery documents from there, and they’re always published while the plugin is active — there’s no toggle to turn them off. You can open each in a browser; they’re JSON.

discovery.json

https://yoursite.com/.well-known/discovery.json is the single, predictable entry point that ties everything together. It’s a structured summary of your site that names:

• your site identity (name, URL, description, language, logo);
• the person or organisation behind it, from your Identity settings;
• links to your other machine documents (sitemap, robots, feed, llms.txt and llms-full.txt when enabled, and more);
• an index of every /.well-known/ document your site serves;
• any APIs, agents, capabilities and resources that Agentimus or another plugin has registered;
• a small trust block (whether the documents are signed, links to your security and privacy policies).

Nothing here is typed by hand — every field is generated from your settings and from what’s actually running on your site, then cached for about an hour. Agentimus also advertises this document in every front-end response so agents can find it from a single link.

agent-card.json (and agent.json)

https://yoursite.com/.well-known/agent-card.json is an A2A (“agent-to-agent”) card — a standard description of your site and any agents it exposes, with your name and organisation as the provider. The same document is also served at /.well-known/agent.json for agents that look there first. On a typical content site the card describes your identity; it grows richer automatically if a plugin registers agent capabilities.

mcp.json

https://yoursite.com/.well-known/mcp.json is an MCP (Model Context Protocol) manifest — the experimental, forward-looking description of any tools an AI agent could use on your site. Agentimus does not run an MCP server itself; it discovers and advertises servers and tools that compatible plugins (for example a WooCommerce or Fluent Cart MCP adapter) have registered. On a site with no such tools, the document is still served but simply reports that none are available yet. Because MCP discovery is still an evolving standard, this document is marked experimental and is kept deliberately separate from the stable discovery.json.

The other /.well-known/ documents

For completeness, Agentimus also serves openapi.json (a machine description of WordPress’s public read-only REST API) and api-catalog (a standard index pointing agents at your API descriptions). Like the three above, these are served automatically while the plugin is active. A few more documents appear only when you opt into the feature that produces them — for example security.txt and tdmrep.json (covered on the crawler-policy page).

The XML sitemap — an opt-in fallback

A sitemap lists your URLs for crawlers. Agentimus deliberately does not compete for this job. It checks, in order:

1. Does WordPress core provide a sitemap? (On by default since WordPress 5.5, at /wp-sitemap.xml.)
2. Does a major SEO plugin provide one? (Yoast, Rank Math, All in One SEO, SEOPress or The SEO Framework.)
3. Only if neither does, and you’ve left the Sitemap (backup) toggle on, Agentimus generates its own at /agentimus-sitemap.xml.

This means Agentimus never produces a duplicate or conflicting sitemap. On most sites something else already owns the sitemap, so Agentimus simply links to that existing file from llms.txt, robots.txt and the discovery document rather than making its own. Its fallback only fills a genuine gap.

When Agentimus’s own fallback is active, it’s a proper sitemap index at /agentimus-sitemap.xml with paginated sub-sitemaps — the same shape core and the SEO plugins use, so even large sites are covered. Password-protected posts are excluded, matching how core’s sitemap behaves.

When it’s served: only when nothing else provides a sitemap and the toggle is on. If you activate an SEO plugin later, Agentimus automatically steps aside and its fallback address stops responding.

Turn it off: Settings → Discovery → untick Sitemap (backup).

A note on privacy and what’s never exposed

Across every file on this page, Agentimus applies the same rule: it only ever exposes what your public HTML site already shows. Draft, pending, private and password-protected content is always excluded, and only the content types you’ve selected on the Content settings appear. Turning a content type off removes it from llms.txt, the full-text edition, the Markdown addresses, the sitemap and the discovery document all at once.

Letting another plugin take over

If your theme or another plugin already produces one of these surfaces — its own llms.txt, its own Markdown delivery, and so on — Agentimus is built to step aside cleanly rather than fight it, and a developer can hand any individual surface over with a single line of code. For most site owners this is automatic and invisible; you’ll only notice that Agentimus never duplicates a file something else is already serving.