Agent Native WP

Provides AI agents (ChatGPT, Claude, Perplexity, etc.) with a clean, structured version of your content—while human visitors continue to see your regular theme as usual.

A WordPress plugin that identifies well-known AI agents based on their user-agent string and delivers a streamlined, machine-readable version of individual pages and posts—as clean, semantic HTML with JSON-LD or as Markdown. Human users continue to see your usual design. It also includes an llms.txt index, rate limiting, and an access log with CSV export.

Features

• Agent Detection - detects known AI agents based on the user-agent string using a configurable list.
• Structured Output - streamlined version per page: clean HTML + JSON-LD or Markdown.
• Format per Request - append ?format=md or ?format=html to each post URL.
• No Impact on Humans - regular visitors will continue to see your theme without being redirected.
• Discovery - /llms.txt (index) and /llms-full.txt (full text), plus a note in robots.txt.
• Clean linking - canonical and <link rel="alternate" type="text/markdown"> on every agent page.
• Rate limiting - only for real agents, fixed time window per IP, HTTP 429 with Retry-After.
• Logging & Monitoring - Dedicated DB table with a dashboard and GDPR-compliant IP handling (masked / hashed / full / none) as well as an automatic retention limit.
• CSV Export - of the logs, with UTF-8 BOM for Excel, batch streaming, and protection against formula injection.

Installation

1. Upload the plugin ZIP file under Plugins → Install → Upload Plugin.
2. Activate—this creates the log table and flushes the rewrite rules for /llms.txt.
3. Configure under Settings → Agent Native.
4. Monitor under Tools → Agent Native Log.

VUsage & Testing

Force agent view in the browser:

?agent_view=1     Append to a post URL
?format=md        Markdown version

Test using a real user-agent:

curl -A "ClaudeBot" https://your-domain.tld/your-page/

Trigger a rate limit:

for i in $(seq 1 100); do
  curl -s -A "ClaudeBot" -o /dev/null -w "%{http_code}\n" https://your-domain.tld/your-page/
done

Check Discovery Endpoints:

https://your-domain.tld/llms.txt
https://your-domain.tld/llms-full.txt

Notes & Limitations

[!IMPORTANT] Show agents the same content as humans, just formatted differently. Displaying different content would constitute deceptive cloaking and may result in penalties. The canonical tag therefore always points to the human-readable URL.

[!NOTE]

• User-Agent detection is spoofable and incomplete—it’s a best-effort heuristic.
• Rate limiting and IP use REMOTE_ADDR. If behind a CDN or proxy, derive the real client IP from a trusted header if necessary (deliberately not the default, as it can be spoofed).
• The Markdown converter is intentionally lightweight (standard tags); for complex Gutenberg blocks, HTML mode is more reliable.

Changelog

1.0.0

• CSV export of the access log (Tools → Agent Native Log).
• UTF-8 BOM, memory-safe batch streaming, protection against CSV formula injection.
• Per-request format (?format=md / ?format=html).
• /llms-full.txt full-text export + robots.txt integration + alternate link.
• Rate limiting for real agents (429 + Retry-After).
• Logging, including a monitoring dashboard and IP privacy modes.
• Initial release: Agent detection, clean HTML/JSON-LD, /llms.txt.

License

Released under GPLv2 (or later).