An AI SEO agent for Claude Code and Cursor
It reads your repo, then names the searches your buyers run.
Each one sized, checked against your own Search Console, and shown with who ranks for it now.
From $49/mo · 7-day free trial · cancel anytime
The real answer a live KDP tool got on 2026-09-23. Its own numbers, not a mockup.
On a live site
Every search you chose, tracked on your own Search Console.
This is the plan board for AppScreenshotStudio, a SaaS we run ourselves, read from production. Each declared search, where it stands, and whether it moved.
How the plan is doing
21 May to 25 Sep · arrows are places moved, up is better
On page 1
where clicks happen · was 0 on 21 May
Below page 1
ranking, earning almost nothing · was 2 on 21 May
Not ranking
no reading on 10+ impressions · was 55 on 21 May
Coverage
<1%The whole site earned 528 clicks in 28 days, most of it on searches this plan never claimed.
On page 1
figma app store screenshot template
#9.9
▲3
320/mo
Demand
38
Impressions
2
Clicks
One line. That's the whole setup.
Run it from the repo of the site you want to work on.
claude mcp add --transport http queryscope https://mcp.queryscope.dev/mcpWorks the same in Cursor, Cline and Windsurf. Your IDE opens a browser once to authorize.
Then ask: “what should this site go after?”
The difference
An SEO agent that works from your code, not from a keyword you type.
Every keyword tool starts with a seed you type, so you get variations of the term you already thought of. QueryScope starts from what your repo says you sell: your agent writes the searches a buyer would type, and QueryScope sizes each one and reads who holds it.
A keyword tool
A word you could think of. So a word you already knew.
Read out of your repo
Searches you would not have typed. Because you did not know them.
Finding the searches is where it starts. Holding them is the part that takes a year.
So the same loop carries it through, in four steps.
It reads your repo, so it knows what you sell.
Your landing page, pricing and positioning, read once and remembered. Without it, rivals get inferred from whoever ranks for the same words.
who buys it · who it is not for · the rivals that matter
It names the searches worth winning.
The searches a buyer types for what you sell, each with its monthly demand, where your own Search Console has you, and who ranks above you.
how big it is · where you stand · who holds it
One target at a time, never a backlog.
One search, the page that owns it, the move, and who ranks above you. Deciding the order is what the product is for. Your agent has the repo open, so it ships it.
the search · the page · the move · who is beating you
Then it says whether it worked.
The clock starts when Google recrawls, not when you shipped. What gets frozen is what the numbers show, not an opinion either of you had.
improved · flat · declined, from your own history
Every change it hands back is one of seven moves, from getting a page indexed to building the page your plan says should exist. See all seven
The plan, then the measurement
Nothing reaches your readers without your commit.
The plan lives at queryscope.md in your repo. Before QueryScope changes a title, a paragraph or the page a search points at, it writes what it intends to do into that file and stops. You read it as a diff, in the tooling you already trust, and your commit is what agrees to it.
Delete a line you disagree with and that counts too: it is recorded as declined, with your reason, and the same proposal is never offered back.
Indexing and internal links run without asking. That is plumbing, and nobody needs a review to fix a broken canonical.
Every change gets a clean measurement.
When your agent records a change, QueryScope starts a clock on it. It counts from the day Google recrawls the page rather than the day you deployed, and it runs at least four weeks of settled data before it will call the result. The result is read against what the rest of your site did over the same days, so a site-wide swing never counts as your change.
-
you ship
day 0
clock not running
-
Google recrawls
the start
before and after now compare
-
settling
4 weeks min
target not handed back
-
verdict
the data calls it
a range, beyond what the site did
A second edit inside that window overwrites the baseline the first one is judged against, so you would lose both. There is no override.
Before each edit, your agent writes down what it expects. The close says whether that held. Over time you get a record of which changes work on your site, kept day by day past the 16 months Google keeps.
Decide it, work it, measure it. On repeat, where you already work.
See pricingPricing
Every feature in every plan. Pick by how many sites you run.
One site, three, or up to 10. Ask it as many questions as you like: the data comes from your own Search Console, so there is nothing for us to meter.
Every plan includes
- Works in Claude Code, Cursor, Cline and Windsurf
- One next move at a time: the search, the page, and who ranks above you
- A plan in your repo, agreed by your own commit
- A brief before you write a page: the search, the reader, what already ranks
- Your clicks, searches and rank, in your editor
- A measurement clock on every change, started when Google recrawls
- The results page read for you below position 10, where Search Console goes blank
- Your history kept longer than Google keeps it
7-day free trial · nothing due today · cancel in one click
FAQ
The questions worth asking first.
How is this different from a keyword tool?
A keyword tool starts from a seed you type, so it can only return variations of a term you already thought of. QueryScope starts from what your repo says you sell: your agent writes the searches a buyer would type, and QueryScope sizes each one and shows who holds it. It then carries the searches you pick through to the change and the measurement, which a keyword tool stops well short of.
How does it know what my site sells?
Your agent reads it: the landing page, the pricing page, any positioning or brand doc, whatever is in the repo. QueryScope records that as the site brief. From the same read your agent proposes the searches your buyers type, and the brief tells QueryScope which rivals to compare you against. Without it, competitors get inferred from whoever ranks for the same words, which finds writers rather than the businesses you lose deals to.
Does it find me new searches to target?
Yes, and it is the first call on any site. Your agent proposes searches from the repo and QueryScope sizes each one. It also lists the searches Google already shows you for that no plan claims (on page 1, at 11 to 30, and at 31 to 100) and the buyer-intent searches a rival holds close to your plan, each with monthly volume attached. You do not need to bring a keyword list.
Does it decide which searches are my buyer?
No, and it says so in its own output. It knows how big a search is, who ranks for it and whether the phrase carries buying intent. It cannot know whether you sell what that searcher wants, because that answer is in your repo. So it names and sizes the candidates, groups the traffic you already have by the job behind each search, and hands the judgement to your agent, which has the repo open.
Does it show me Google's AI answers?
Yes, for the searches in your plan. Each results-page reading shows whether Google gave an AI answer and whether it cited you. It also names the sites it cited instead. Across your plan, your agent sees how many answers cite you now and a month earlier. It also sees which sites get cited on several of your searches. Each reading is one market on one day, so one search changing is a single sample. It reads Google's answers only, not ChatGPT or other chat apps.
Does QueryScope change my code?
No. QueryScope never sees your repository and never writes a file. Your own coding agent makes every edit, in your editor, under your permissions. For anything that changes what a visitor reads, QueryScope writes the proposal into queryscope.md first and waits for you to commit it.
Why will it not give me the next task right away?
Because a second edit inside a measurement window destroys the baseline the first one is judged against, and then neither can be judged. Once you record a change, that target waits until the clock is up. Everything else in your plan is still available to work on meanwhile.
Is this connected to Google?
QueryScope is independent and not affiliated with Google. It reads your own Search Console over a read-only grant you approve, and you can revoke it at any time from your Google account or from ours.
Search Console is free. Why pay for this?
Search Console shows you numbers. It does not hold the plan, decide what to do next, read the results page below position 10 where it goes blank, or tell you whether last month's change worked. It also deletes everything past 16 months, and we keep your history past that.
Which editors work?
Claude Code, Cursor, Cline and Windsurf. It is a standard remote MCP server over Streamable HTTP, so any client that supports remote MCP can connect with the same line.
Will it tell me if the extra clicks turn into signups?
No, and it will never claim to. QueryScope stops at the click: it reads search data and nothing past it. Whether those visitors convert is a measurement in your analytics or your product, and any tool telling you a ranking change earned you revenue is guessing.
Add one line. Ask what to work on.
Two minutes to set up. Then your editor tells you which search to win next, and what is standing in the way.
7-day free trial · nothing due today · cancel in one click