Your buyers already told you what your site should say, on the phone and in chat, and nobody had time to read it back. Pain to SEO reads that pile, finds the questions your site never answers, checks that the page does not repeat one you already have, writes it against the same rules our own sites are held to, publishes it as a draft, and then opens the live address to make sure it is not broken. Everything runs on your own machine.
In plain words
Your customers already told you what your website should say. It is sitting in your call recordings, your chat history and the notes you type after a call, and nobody has ever read it back because there is far too much of it. This reads the whole pile for you, finds the questions people keep asking that your site answers nowhere, and then takes the best one all the way: it checks the subject will not clash with a page you already have, gives your own AI assistant everything it needs to write the page properly, puts it on your site as a draft for you to approve, and finally opens the published page to make sure it actually looks right to a visitor. Your conversations and recordings never leave your computer.
Who it is for
An owner who has answered the same question on the phone fifty times and never thought to put it on the site
A marketer who needs a list of topics that is not guesswork, and the finished pages to go with it
Anyone paying for content who wants the order decided by customers rather than by a content plan
Anyone who has published a page and only found out later that it was broken on the live site
What you actually do
Put your call recordings, chat exports or notes in a folder
Ask your assistant, in your own words, to find what your site is not answering
Read the shortlist and pick the question worth a page
Approve the draft it writes and let it publish and check the result
$69
One licence per owner, any number of sites and archives. Conversations and recordings stay on your machine: no upload, no account, no telemetry
Payment is taken by Whop. Buying means you accept the terms and the
refund policy: what is included, when it arrives and when we give the money back.
What you need before you start
What
Cost
Why
Node.js 20 or newer
Free
Everything runs on your own computer, which is why the conversations never leave it. Node.js is free software from nodejs.org.
An AI assistant you already use
You pay your own assistant, nothing extra to us
Claude, Cursor or anything that speaks MCP. It does the reading and the writing with your facts; we give it the method, the rules and the tools. Nothing about your business passes through us.
Conversations to read
Free
Exports from any support tool or CRM, or call recordings, or notes in a document. Under about a hundred exchanges the counts mean little and you would learn more reading them yourself.
A way into your site, only if you want it to publish
Free
A WordPress application password, or the folder of a site built from files. Without it you still get the finished text and paste it yourself.
A browser, only for the readability check
Free
One optional command installs it. Without it every other check still runs and the report says plainly that contrast was not measured, rather than implying the page is fine.
What you get
Your calls turned into text on your own machine, free, with no recording sent anywhere
The questions your buyers actually ask, merged when they are the same question, with how many times each came up
For each question: whether your site answers it, mentions it in passing, or says nothing at all, with the closest page and the paragraph it found
A collision check before you write: whether an existing page already claims that subject, and which sentences a draft is repeating from your own site
A brief for whoever writes: the question, the numbers, the neighbouring pages and the rules the page has to pass
The same content rules our own sites are held to, in the package: no empty sections, no title cut in the middle, no figure without a source
Publishing that cannot embarrass you: WordPress gets a draft with an edit link, a file-built site gets one file matching its neighbours, and nothing goes live on its own
A check of the published page as a reader sees it: text rendered, images loading, links alive, not hidden from search, contrast readable
How it works
1. Bring what you already have. Call recordings, a chat export, support tickets, a WhatsApp thread, notes typed after a call. Recordings are transcribed on your own computer, free, and are never uploaded.
2. Hear the question under the words. It keeps the customer side and drops your own replies, then merges the same question asked five different ways and counts how often it really came up. The order is set by your buyers, not by a content plan.
3. Hold it up against your site. Every question is looked for on your own pages. You get three piles: nothing on the site answers this, it is mentioned in passing, this is already covered. The passages are shown so you judge, not a score.
4. Check it will not fight your own pages. Before a word is written: does an existing page already claim this subject, and which sentences would be repeated from your site. This is what stops two pages competing for the same search and splitting it.
5. Write it properly. Your assistant gets the brief and the rules: answer in the first paragraph, your real numbers, your real prices, sources for figures, the structure that search engines and AI assistants both read. Your assistant writes, on your machine, with your facts.
6. Put it on the site. WordPress gets a draft and a link to open it. A site built from files gets one file in the right folder, with the same frontmatter its neighbours use. Nothing is published live and nothing is pushed without you.
7. Open it and make sure it is right. The last step visits the published address and checks what a build log cannot see: that the article really rendered, that every image loads, that links go somewhere, that the page was not left hidden from search, and that the text is readable. If anything is wrong, the job is not done.
What it is not for
Anyone with no archive of customer conversations yet: there is nothing to read
Anyone expecting a machine to decide what is true: it shows the passages and the counts, you and your assistant judge
Tilda, Wix and other closed builders: there is no honest way in, so the last step is you pasting the finished text
Anyone wanting it published live untouched: pages go up as drafts on purpose, a person presses publish
The sites in this curve are the ones this toolkit was built for and tested on.
Why I share this
I sell property to international buyers. It is a niche where a single advertising lead costs more than most products on this page, and where the buyer reads for weeks before writing to anyone. For years the enquiries came from paid ads. Then I built content sites that answer the questions those buyers ask, and the enquiries started arriving on their own: from Google first, then, more and more, from ChatGPT and Perplexity, which quote the pages when someone asks.
The numbers are on the cases page, site by site: clicks from Google up 249 percent across thirteen sites in more than ten countries, the first 28 days of June against the 28 days to 25 September 2026, and a new site that went from 3 clicks in its first 28 days to 98. Every tool on this page is a piece of the pipeline behind those sites, stripped of my data and packaged so you can run it in your own niche: the method for finding what a market actually searches for, the page shapes that get quoted, and the checks that stop a bad page from going live.
I am not a software company. I am an operator who got tired of paying for every lead, and these are the tools I use every week. If they work for you the way they work for me, tell me; if they do not, the refund policy says exactly what comes back and when.
Do my conversations or call recordings get uploaded?
No. Recordings are transcribed on your own computer and the files are read there. The only request that leaves the machine goes to your own website, which is what a search engine does anyway. There is no account, no key and no telemetry in the package.
Who actually writes the page?
Your own assistant, on your machine, with your facts. We give it the brief, the context from your site and the rules the page has to pass. That is deliberate: you pay your assistant directly, nothing about your business passes through us, and nobody is charged twice for the same words.
Will the new page compete with one I already have?
That is checked before anything is written. The kit looks for an existing page that already claims the subject, and for sentences the draft would be repeating from your own site. Two pages chasing one search split it between them, and this is what stops that happening.
Can it publish for me?
It can put the page on the site, always as a draft. WordPress gets a draft and an edit link; a site built from files gets one file in the right folder matching its neighbours, with nothing committed or pushed. A page written by a machine goes past a person first.
What is the check after publishing for?
A page can be written well, pass every rule, be published and indexed, and still be broken for whoever opens it: an image the CDN refuses, text on a background it was never coloured for, a link to a page that no longer exists. A build log sees none of that. The last tool opens the live address and looks, so nobody reports the job done on a broken page.
What formats does it read?
Plain text, Markdown, CSV, TSV, JSON and log files, from a folder or a single file, plus common audio and video for calls. No particular export shape is required beyond one line per turn.
Does it work in Russian?
Yes, and in English. Both are handled in the same run: it detects which side of a conversation is the customer in either language.
Do I need to use a terminal?
No. It runs as an MCP server inside Claude or Cursor, so you ask your assistant in your own words and it calls the tools. There is a one-line terminal command too, if you prefer it.
OperStack