Kafkai Documentation

Kafkai connects the AI you already use to real search data. Ask about any site, your own or one you are curious about, and get the answer in numbers: where it ranks, who links to it, who it competes with, and what to write next.

Kafkai analyses your site and your competitors' sites, classifies the keyword landscape with the patented 4C framework, and hands the results to the AI agent you already use: Claude, ChatGPT, Claude Code, or any other client that speaks the Model Context Protocol (MCP).

What Kafkai is for

You ask your agent about a site, and the agent answers from data that Kafkai supplies. When you run your own site, the questions repeat: what to write or fix, and what the change did after you published. Kafkai covers the data and the arithmetic. Your agent covers the thinking and the writing.

  • Any site. Rankings, keywords, backlinks and competitors work for any domain: your own, a competitor's, a client's. Search Console and analytics data work only for a domain you have verified.
  • Data, then a decision. Every keyword comes back as one of four moves: compete, catch up, consolidate, or complement. Your agent turns those moves into what to write and what to fix.
  • Memory. A project keeps what Kafkai recommended, what your agent published through it, and how rankings moved afterwards. Every rank Kafkai measures is dated, so your agent can ask what changed since the last update, or over the last three months.
  • Your stack, not ours. Kafkai has no LLM of its own. Your agent thinks and writes with the model you already use, and Kafkai sits above the model layer. Switching models later does not break the project.

Why your agent needs Kafkai

An LLM analyses well and calculates badly. Ask it for the pages with the best click-through rate over the past four weeks, and something has to hold four weeks of Search Console data, compute the rate per page, sort the result, and hand back a list. The model on its own has neither the data nor reliable arithmetic.

Kafkai does the arithmetic. Your agent does the thinking.

  • Kafkai holds the data, runs the computation, and returns structured numbers.
  • Your agent interprets the numbers, decides, writes, and presents the result in whatever form you want: a report, a table, a chart.

How it works

  1. You create a project for your site, and add competitor sites if you have any. Competitors are optional. Without them Kafkai still reads what your own site ranks for, and you can add competitors later.
  2. Kafkai fetches search ranking data and classifies every keyword into one of the 4C strategies: Compete, Catch-up, Consolidate, and Complement. For more information, see the 4C framework page.
  3. Your agent connects to the Kafkai MCP server and reads that data through tools. It then researches, drafts, and revises on your side of the connection.
  4. After you publish, your agent asks Kafkai what moved: rankings, Search Console queries, and backlinks, all dated.

What Kafkai does not do

  • Kafkai does not write articles. Your agent writes with its own LLM, and Kafkai keeps only a snapshot of what it published.
  • Kafkai does not sell rank tracking. Rankings refresh when you ask for an update, not on a schedule.
  • Kafkai does not promise rankings or traffic. It gives your agent the same picture of the search landscape that an editorial team would build by hand, and then measures what happened.

What that looks like in a conversation is on the Example Prompts page: a prompt, the tools the agent calls behind it, and what comes back.

The MCP server

Your agent connects to one endpoint:

https://kafkai.com/mcp

Authentication is either an OAuth 2.1 sign-in (interactive clients) or a static API token (headless clients). For more information, see Authentication.

Contents

About the 4C framework

The 4C framework — Compete, Catch-up, Consolidate, Complement — is covered by JPO Patent No. 7701764. KAFKAI® is a registered trademark (Reg. No. 6661421).