The analysis runs before anything is installed. You see the 23 checks on your pages first, from outside, in a free account. Installing the agent is what lets it act on what it found — one file on a coded site, three steps on WordPress.
Put your domain into the free check and the agent reads your public pages from outside: 23 structural checks, the evidence behind each one, and what to change, in order. A free account, no card, nothing added to your site. You see what it found before you decide anything.
Next.js, Astro, a static build, anything behind a CDN. Drop
reflexa-agent.js into your repository — one file, no dependencies,
no build step — and wrap your handler with it wherever every request already
passes: a Cloudflare Worker, Next.js or Netlify middleware, your own Node
server.
From there it does what the plugin does, minus the parts only a CMS can do.
It counts every AI visit page by page; it serves /llms.txt when you
switch that on; it appends the AI-fetcher rules to your own
robots.txt; and it injects Organization data into
<head>. Each of those is a switch in the dashboard and takes
effect on the next request.
What it does not do, and says so. It never writes to your codebase, never touches your build, and never changes a page's words — on a coded site the text lives in your repository, and that is where it should change. The agent writes the exact words and tells you which file they belong in. The sitemap switch is not offered either: a sitemap lists your own URLs and your build owns it.
The copy-paste snippet, with your token already in it, is on the Reflexa agent page in the dashboard.
No developer, no FTP, no theme edits. The plugin does not write into your theme or your content: the switches live in its own options and the page edits land as ordinary WordPress revisions you can undo.
Its own version, what it is running on, and the list of pages the site publishes. On WordPress it also reports the WordPress version and active plugins, the SEO plugin in use, whether the site is set to discourage search engines, and the permalink structure. It reports those so the dashboard can stop guessing about your site — and so the agent's answers are about your actual setup.
It does not read customer data, orders, or anything behind an account. The full list of what it never does →
On a coded site, remove the one file and redeploy; whatever was last switched on stops being served on the next request. On WordPress, deactivate the plugin. The switches stop being applied; turn each one off first if you want the site reverted exactly. Page edits were revisions, so they restore like any other revision. Your checks and the agent's memory stay in your Reflexa account, waiting, in case you come back.
A coded site needs someone who can add one file to your middleware, CDN worker or template — the dashboard gives you the exact snippet, with your token already in it. On WordPress, no: install the plugin, paste the token, done.
Access to that one site's own checks and nothing else. The server refuses the token if the plugin reports from a different host, so two sites on one account stay completely separate.
No. The plugin answers Reflexa on its own schedule and the counting adds nothing to the page a visitor loads.
Yes — that is the default. Put in your domain and the free check reads your public pages from outside.