Documentation
Everything you need to connect That SEO Agent and use it well
There is no separate app to learn. You run the server on your own machine, add one line of configuration to the AI assistant you already use, and then ask it questions about your site in plain language. These pages cover that setup, what each part does, and the ideas behind how we measure things.
Start here
Three pages, in the order most people need them. If you only read one, read the first.
Connect your AI assistant
The setup, start to finish: clone and start the server, create your own Google OAuth client, and point Claude Code or Cursor at the loopback address it prints.
How to use it day to day
What to actually ask once it is connected, and the kinds of questions it answers well.
How it works
What happens between your question and the answer, and where the data comes from.
Reference
The detail behind the setup, for when something does not behave the way you expected.
The MCP server
What the server exposes, how it authenticates, and the tools it offers your assistant.
Search Console over MCP
Which Search Console data is reachable, and the shape of the answers you get back.
Our crawler
How our crawler identifies itself when it reads a page, and how to allow or block it.
Concepts
The reasoning behind what we measure. Worth reading before you act on a score, ours or anyone else's.
Generative engine optimization
What changes when the thing reading your site is a language model rather than a search index.
How we score AI visibility
What our score is built from, and the limits of what any score like it can tell you.
What we do not do
The things we deliberately leave out, and why leaving them out is the honest choice.
For agents and crawlers
If you are software rather than a person, these four files describe the whole site without reading any of the pages above.
Credentials and access
The open-source server needs no key from us. What it needs is your own Google OAuth client, created in Google Cloud Console and held in a `.env` file next to the server; both scopes it requests are read-only, and the tokens it stores never leave your machine.
How to set it upThe full tool list
We do not publish a hand-written page listing every tool. Your assistant asks the server what it can do the moment it connects, and the server card carries the same list, both generated from what is actually running. A page copied by hand would be out of date within the month, and a list that is wrong is worse than no list.
Versioning and deprecation
What counts as a breaking change, the notice you get before one, and what we promise not to do. Published as plain markdown so your assistant can read it without you.
Read the policySomething missing
If a page here did not answer your question, that is worth knowing about. Tell us what you were looking for.
Get in touch