What's working for that site?
Ask your AI.
Where a site ranks, who links to it, what to write next. Connect Kafkai to Claude, ChatGPT or your own agent, and ask about your own site or anyone else's. The answers come back as real data.
Start with a question
you already have.
Type it in plain words, about your own site or any other. Below are six questions and the answer that comes back. For more, copy a prompt from the Example Prompts page and change the domain.
“Which keywords does example.com rank for?”
Every keyword the site ranks for, with its position and monthly search volume.
“Who does my site compete with in search?”
Sites that rank for the same searches, how much their keywords overlap with yours, and whether each one is a direct competitor or a portal site.
“Which sites link to my competitors but not to me?”
The sites with a live link to your competitors and none to you: the list an SEO tool calls a link intersect, ready for outreach.
“What should I write next?”
The keywords your competitors rank for and you don't, the content gap, with search volume, so your AI can pick the first articles to write.
“Which of my pages are losing clicks?”
Pages and searches whose clicks dropped, from your own Search Console. This works for your own site once you verify the domain.
“How did that site's rankings change since the last research?”
Keywords that moved up, moved down, appeared or dropped out between two research runs.
Kafkai is a plugin
for the AI you already use.
You ask in your AI, and the answer comes back in your AI. Kafkai works in between, so you never open a Kafkai screen to ask. The technical name for this kind of plugin is MCP. Setup is a one-time job of about 30 minutes.
ASK
You
Type a question in Claude, ChatGPT or your own agent, in plain words.
CALL
Your AI
Your AI sees that the question needs search data, and calls the Kafkai tools that fit.
LOOK UP
Kafkai
Kafkai reads real search data: rankings, backlinks, competitors, and Search Console for a site you have verified.
CALCULATE
Kafkai
Kafkai counts, sorts and compares, and returns the numbers to your AI.
ANSWER
Your AI
Your AI reads the numbers and answers the way you asked: a list, a table, a chart or an article draft.
Same AI.
Different answer.
Ask your AI about a site on its own, and it guesses from what it has read on the web. Connect Kafkai, and the same AI answers from real search data: the keywords the site ranks for, the sites that link to it, and the sites it competes with.
An AI cannot make that data, and it calculates badly. Ask for your best articles by CTR over the past four weeks, and something has to hold the data, do the arithmetic and sort the list. Kafkai does that part. Your AI reads the result and decides what it means.
What changes is also what you know. The question you had about a site stops being a question. It becomes a list with numbers next to it, and you decide from that.
From guesses to numbers
Which sites link to a competitor? Without Kafkai, your AI guesses. With Kafkai, it gets the list.
Featured by leading media
Any site. Real numbers.
Pay only for what you ask.
Three ways people use Kafkai, the patented method underneath, what it costs, and what it does not do.
Any site
Your own site, a competitor, a client, or a site you saw yesterday. A quick look at one site is a normal use, not a trial. One project with one competitor fits in the smallest pack, ¥1,100 including tax.
Data, then a decision
Without Kafkai, your AI guesses about a site from what it has read. With Kafkai, the same AI answers from real search data. It turns the numbers into what to write, what to fix and why, and shows the numbers behind each point.
Ask, write, publish, check
Your AI writes the article. Kafkai sends it to your WordPress site as a draft, so you check it before it goes live. Later, ask what happened: ranking changes from the next research run, and Search Console clicks once the domain is verified.
The 4C framework
Compete, Catch-up, Consolidate, Complement: the four kinds of recommendation Kafkai sorts a site's keywords by comparing them with its competitors. Fight head-on, catch up, merge, or add next. Registered with the Japan Patent Office as Patent No. 7701764.
Pay for what you use
No monthly subscription. A session that researched a site and its competitors came to ¥1,983. A session that only read existing data came to ¥62. Your AI writes on the plan you already pay for, so you don't pay for AI writing twice. There is no monthly fee to justify before you know whether the data helps.
What we don't do
We don't watch rankings every day or on a schedule. Research runs when you ask for it. We don't promise rankings or traffic. We don't write your articles, and we don't store them. Your AI writes; your site keeps the article.
For people who run a site.
And the teams who run many.
Small companies without an SEO person, people who run a site on their own, agencies with client portfolios, and web production companies with maintenance contracts.
Site owners
A small company or one person with a site, often with no SEO person. Ask about your site and the sites around it in plain words. After the first answer you know which keywords the site next to you wins, and what to write stops being a guess.
Marketing agencies
Several client sites, one question each. Ask for the month's changes per client and the numbers come back ready for the report, so the status call starts from a list, not from a morning of copying screens.
Web production companies
Your maintenance contracts include content. Every retainer gets a monthly, numbers-backed answer to “what should we do next”, and the client hears a list instead of a hunch.
Frequently Asked Questions
Plain answers — including what Kafkai doesn't do.
Kafkai connects the AI you already use, such as Claude, ChatGPT, Claude Code or your own agent, to real search data. Ask about any site, your own or one you are curious about, and get the answer in numbers. It connects to your AI as a plugin; the technical name is MCP.
No. Your own AI does the writing, in your voice, on the plan you already pay for. Kafkai supplies the data and the arithmetic underneath, and can send what your AI wrote to your site as a draft.
On its own, ChatGPT has no search data about a site. Ask it where a site ranks or who links to it, and it guesses from what it has read on the web. Connect Kafkai, and the same AI answers from real data. Kafkai also does the arithmetic an AI does badly, such as sorting your articles by CTR over four weeks. An AI cannot make that data itself. You can test this: ask the same question with Kafkai off, then on.
Kafkai shows where a site ranks, and how that changed between research runs, when you ask. It does not watch rankings for you every day, and it does not promise rankings or traffic. Research runs only when you start it, because every run costs credits. For your own site, Search Console shows what actually happened: impressions, clicks and CTR.
Compete, Catch-up, Consolidate, Complement — the four kinds of recommendation Kafkai computes: where to fight head-on, what to catch up on, what to merge, and what to add next. The framework and the underlying competitive-analysis technology are covered by JPO Patent No. 7701764, granted 24 June 2025; KAFKAI® is trademark Reg. No. 6661421.
Prepaid credits: a fixed charge per call plus a smaller charge per data row returned. No monthly subscription. Buy credits, use them, and top up. As an example, a session that researched a site and its competitors came to 1,983 credits (¥1,983), and a session that only read existing data came to 62 credits (¥62). Every new account starts with 1,000 free credits.
Anything that speaks MCP: Claude, ChatGPT, Claude Code, Cursor, or an agent you built yourself. Switching models later doesn't break anything: your projects, research and history stay.
Add Kafkai as a custom connector. In claude.ai open Customize, then Connectors, choose Add custom connector, name it Kafkai and enter the server URL https://kafkai.com/mcp. Claude opens a Kafkai sign-in page: sign in and approve the access request. In a chat, turn the connector on from the + menu and ask "List my Kafkai projects" to check that it works.
The same connector reaches Claude Desktop and Claude Code on the web (claude.ai/code), so there is nothing more to set up there. On Team and Enterprise plans an organization owner adds the connector first. Step by step: Connect Claude. What to ask once it is connected: Example Prompts.
If you sign in to Claude Code with your claude.ai account, the Kafkai connector you added in claude.ai is already available, in the terminal and in Claude Code on the web. Otherwise add the server from a terminal:
claude mcp add --transport http kafkai https://kafkai.com/mcp
Then run /mcp inside Claude Code, or claude mcp login kafkai from the terminal, and approve the sign-in in your browser. Add --scope user to the first command to make Kafkai available in every project. Step by step: Connect Claude. What to ask once it is connected: Example Prompts.
Hermes runs without a browser, so it authenticates with an API token instead of a sign-in. Create a token on the API tokens page, then add Kafkai under mcp_servers in ~/.hermes/config.yaml with the server URL https://kafkai.com/mcp and the header Authorization: Bearer YOUR_API_TOKEN. Run /reload-mcp in Hermes and the Kafkai tools appear with the mcp_kafkai_ prefix.
Any other MCP client that can send an Authorization header connects the same way. The exact YAML, and how to keep the token out of the config file: Connect Hermes and Other Clients. What to ask once it is connected: Example Prompts.
An SEO tool gives you a dashboard, a monthly fee, and the job of reading the reports yourself. Kafkai gives the same kinds of data, competitor keywords, keyword demand, backlinks and what happened after you published, to the AI you already use, and the AI reads them for you. Three practical differences: you pay per question, with no monthly fee to justify before you know whether the data helps; one conversation covers competitor keywords, backlinks, Search Console and analytics, where a suite needs a separate report for each; and there is no screen to learn.
Pick a site. Ask your first question.
Register, connect Kafkai to your AI, and ask about your own site or any other. New accounts start with 1,000 free credits, enough to research your first site with one competitor and read what comes back.