gregrandall

AI agents, LLMs, and crawlers have to wade through navigation bars, sidebars, ads, and comment forms to reach the content they want, and every element costs tokens. <a href="https://blog.cloudflare.com/markdown-for-agents/" rel="nofollow ugc">Cloudflare measured</a> an 80% reduction in token usage when converting a blog post from HTML to Markdown (16,180 tokens down to 3,150).
Botkibble adds a Markdown endpoint to every published post and page.
Cloudflare offers <a href="https://developers.cloudflare.com/fundamentals/reference/markdown-for-agents/" rel="nofollow ugc">Markdown for Agents</a> at the CDN edge on Pro, Business, and Enterprise plans. Botkibble does the same thing (for free) at the origin, so it works on any host.
<a href="https://github.com/greg-randall/botkibble" rel="nofollow ugc">GitHub Repository</a>
Three ways to request Markdown:
<ul>
<li><code>.md</code> suffix: append <code>.md</code> to any post or page URL (e.g. <code>example.com/my-post.md</code>)</li>
<li>Query parameter: add <code>?format=markdown</code> to any post or page URL</li>
<li>Content negotiation: send <code>Accept: text/markdown</code> in the request header</li>
</ul>
What’s in every response:
<ul>
<li>Structured metadata header with title, date, categories, tags, word count, character count, and estimated token count (in YAML frontmatter format, readable by any AI agent)</li>
<li>Clean Markdown converted from fully-rendered post HTML (shortcodes run, filters applied)</li>
<li><code>Content-Type: text/markdown</code> and <code>Vary: Accept</code> response headers</li>
<li><code>Content-Signal</code> header for AI signal declaration — defaults to <code>ai-train=no, search=yes, ai-input=yes</code> — see <a href="https://contentsignals.org/" rel="nofollow ugc">contentsignals.org</a></li>
<li><code>X-Markdown-Tokens</code> header with estimated token count</li>
<li>Discovery via <code><link rel="alternate"></code> in the HTML head and <code>Link</code> HTTP header</li>
<li>Automatic cache invalidation when a post is updated or deleted</li>
</ul>
Performance:
Botkibble writes Markdown to disk on the first request, then serves it as a static file. A built-in Fast-Path serves cached files during WordPress’s <code>init</code> hook, before the main database query runs. No extra configuration needed.
Add a web server rewrite rule and Botkibble bypasses PHP entirely, serving <code>.md</code> files the same way a server would serve an image or CSS file:
 Method 
 Avg. response time
 Standard HTML 
 0.97s
 Markdown (cold, first request) 
 0.95s
 Markdown (cached, PHP Fast-Path) 
 0.87s
 Markdown (Nginx/Apache direct) 
 0.11s
Serving directly from disk is 88% faster than a full WordPress page load. See the Performance section below for Nginx and Apache configuration.
Security:
<ul>
<li>Drafts, private posts, and password-protected content return <code>403 Forbidden</code></li>
<li>Rate limits cache-miss regenerations (20/min by default) to mitigate DoS abuse</li>
<li><code>X-Robots-Tag: noindex</code> keeps Markdown versions out of search results</li>
<li><code>Link: rel="canonical"</code> points search engines back to the HTML version</li>
</ul>
Cache variants (optional):
You can persist alternate cached representations by adding <code>?botkibble_variant=slim</code> (or any other variant name). 
Variant caches are stored under:
<pre><code>/wp-content/uploads/botkibble/_v/<variant>/<slug>.md
</code></pre>
What it does NOT do:
<ul>
<li>Expose drafts, private posts, or password-protected content</li>
<li>Serve non-post/page content types by default</li>
<li>Require any configuration. Activate and it works.</li>
</ul>
<h3>Why Markdown?</h3>
HTML is expensive for AI systems to process. <a href="https://blog.cloudflare.com/markdown-for-agents/" rel="nofollow ugc">Cloudflare measured</a> an 80% reduction in token usage when converting a blog post from HTML to Markdown (16,180 tokens down to 3,150).
Cloudflare now offers <a href="https://developers.cloudflare.com/fundamentals/reference/markdown-for-agents/" rel="nofollow ugc">Markdown for Agents</a> at the CDN edge via the <code>Accept: text/markdown</code> header, available on Pro, Business, and Enterprise plans.
This plugin does the same thing at the origin, so it works on any host. It also adds <code>.md</code> suffix URLs, <code>?format=markdown</code> query parameters, YAML frontmatter, static file caching, and server-level offloading.
If you use Cloudflare, both share the same <code>Accept: text/markdown</code> header, <code>Content-Signal</code> headers, and <code>X-Markdown-Tokens</code> response headers.
Cloudflare currently defaults to <code>Content-Signal: ai-train=yes, search=yes, ai-input=yes</code> with no way to change it. Botkibble defaults to <code>ai-train=no</code> and lets you override the full signal per site via the <code>botkibble_content_signal</code> filter.
<h3>Performance & Static Offloading</h3>
This plugin supports static file offloading by writing Markdown content to <code>/wp-content/uploads/botkibble/</code>.
<h3>Nginx Configuration</h3>
To bypass PHP entirely and have Nginx serve the files (including variants) directly:
<pre><code># Variants
location ~* ^/(_v/[^/]+/.+)\.md$ {
 default_type text/markdown;
 try_files /wp-content/uploads/botkibble/$1.md /index.php?$args;
}

# Default
location ~* ^/(.+)\.md$ {
 default_type text/markdown;
 try_files /wp-content/uploads/botkibble/$1.md /index.php?$args;
}
</code></pre>
<h3>Apache (.htaccess)</h3>
Add this to your <code>.htaccess</code> before the WordPress rules:
<pre><code>RewriteEngine On
# Variants
RewriteCond %{DOCUMENT_ROOT}/wp-content/uploads/botkibble/_v/$1/$2.md -f
RewriteRule ^_v/([^/]+)/(.+)\.md$ /wp-content/uploads/botkibble/_v/$1/$2.md [L,T=text/markdown]

# Default
RewriteCond %{DOCUMENT_ROOT}/wp-content/uploads/botkibble/$1.md -f
RewriteRule ^(.*)\.md$ /wp-content/uploads/botkibble/$1.md [L,T=text/markdown]
</code></pre>
Even without these rules, the plugin uses a “Fast-Path” that serves cached files from PHP before the main database query is executed.
<h3>Credits</h3>
We thank Cristi Constantin (https://github.com/cristi-constantin) for contributing cache variants, URL and SEO improvements, and fixing important bugs.

Botkibble

Serves every published post and page as Markdown for AI agents and crawlers. No configuration, no API keys. Activate and it works.

phalkmin

ParseLess serves your WordPress content as clean Markdown to AI crawlers (via User-Agent detection) and on manual <code>?format=md</code> requests. Same URL, same content, none of the theme chrome, navigation, widgets, or page-builder scaffolding that AI bots and CLI tools don’t need.
It also exposes <code>/llms.txt</code> for the emerging AI-indexing standard, so models know where to find your content.
<h4>Who is this for?</h4>
1. Developers using Claude Code, Cursor, Aider, or any CLI that feeds your own site content into an LLM.
If you’ve ever piped a blog post into Claude or ChatGPT for analysis, rewriting, or summarization, you’ve watched it burn through tokens parsing nav menus, footer markup, and CSS classes that have nothing to do with your content. A typical WordPress page measured live: ~19,800 tokens of HTML for ~975 tokens of actual content — a 20x reduction just by stripping the theme. Heavy page-builder sites (Elementor, Divi) routinely hit 100x or more.
That’s the difference between fitting a handful of pages in a context window and fitting dozens.
Just append <code>?format=md</code> to any post URL and pipe it straight into your tool of choice:
<pre><code>curl https://yoursite.com/my-post/?format=md | claude "summarize this"
</code></pre>
2. Site owners getting “high resource usage” warnings from their host because AI bots are hammering the site.
GPTBot, ClaudeBot, PerplexityBot, Google-Extended, and a dozen others crawl WordPress sites constantly. Each request renders your full theme, runs widget queries, loads page-builder assets, and ships hundreds of KB of HTML per page — most of which the bot discards before extracting the actual text.
ParseLess intercepts these crawls and serves a tiny Markdown payload instead. You keep the AEO/SEO benefit of being indexed by AI search, without paying the server cost of rendering your full theme for every bot hit.
Typical impact on AI bot traffic (measured on real WordPress pages):
<ul>
<li>95%+ less bandwidth per crawled page (typically 15–30x smaller; e.g. a 79 KB page → 4 KB. Heavy page-builder sites see 100x+)</li>
<li>~60–80% fewer database queries per request</li>
<li>~60% lower peak PHP memory per request</li>
<li>~50–80% faster Time To First Byte (measured 54% on a content-heavy page, 67% on a short one)</li>
<li>On a site with 500 posts crawled monthly by major AI bots, that’s roughly gigabytes → tens of megabytes of monthly bandwidth and a fraction of the cumulative PHP execution time</li>
</ul>
Numbers vary by theme and content. Heavier setups (Elementor, Divi, Avada) see the biggest savings; lightweight themes see less dramatic but still meaningful gains. The conversion is cached as a transient, so repeated bot hits cost almost nothing.
<h4>How it works</h4>
<ul>
<li>AI crawlers are detected by User-Agent and served Markdown automatically — no configuration required.</li>
<li>Humans, search engines, and unknown bots receive your normal HTML output. ParseLess never affects what real visitors see.</li>
<li>The <code>?format=md</code> query parameter works on any post URL for manual preview or CLI piping.</li>
<li><code>/llms.txt</code> is published at your site root with a list of available content for AI indexers.</li>
<li>Conversion happens once per post and is cached. Subsequent requests serve a single transient read.</li>
</ul>
<h4>Features</h4>
<ul>
<li>Automatic Markdown for known AI bots (GPTBot, ClaudeBot, PerplexityBot, OAI-SearchBot, CCBot, Google-Extended, Applebot-Extended, Bytespider, Meta-ExternalAgent, cohere-ai, and more)</li>
<li>Manual preview and CLI access via <code>?format=md</code> on any post URL</li>
<li><code>/llms.txt</code> endpoint for AI-indexing standards</li>
<li>Respects noindex flags from Yoast SEO, Rank Math, and Genesis</li>
<li>Skips private, draft, password-protected, and trashed posts</li>
<li>Works with all public post types (configurable)</li>
<li>Optional YAML frontmatter (title, URL, author, date, categories, tags, excerpt)</li>
<li>Transient-based caching with configurable TTL</li>
<li>Settings page at Tools → ParseLess for detection mode, bot list, cache TTL, post types, and llms.txt control</li>
<li>Per-post meta box with Markdown preview and copy-to-clipboard</li>
<li>Extensible via filters: <code>md4ai_bot_list</code>, <code>md4ai_supported_post_types</code>, <code>md4ai_markdown_output</code>, <code>md4ai_cache_ttl</code>, <code>md4ai_should_serve_markdown</code></li>
</ul>
<h3>Privacy and data collection</h3>
When logging is enabled (off by default), ParseLess records the following for 
each Markdown request:
<ul>
<li>Post ID, URL, and request timestamp</li>
<li>The full User-Agent header</li>
<li>A salted SHA-256 hash of the requester’s IP address (the raw IP is never stored)</li>
<li>Matched bot identifier, bytes served, and whether the response came from cache</li>
</ul>
IP hashes cannot be reversed from an email address, so the WordPress privacy 
exporter/eraser tools will report no personal data on demand.
Logs are pruned daily according to the configured retention window 
(7/30/90/365 days, default 30). Site owners can disable logging or click 
“Delete all logged requests” at any time from Tools → ParseLess.

ParseLess

Serve your posts as Markdown to AI crawlers and CLI tools — typically 95%+ less bandwidth, ~60–80% fewer DB queries, no theme bloat.

Botkibble vs ParseLess