Which pages show up, which are wasted. In one call.
Your agent connects to one endpoint and reads the answer from your own Search Console.
- No card. Sign in with Google.
- You need a verified property in Search Console.
- Google shows an "unverified app" screen; verification is queued.
Paid plans are waiting on payment setup; the free tier is open today.
- 1
- call, one answer
- 0.4 s
- to analyse 5,442 pages
- < 900
- tokens, even our largest response
- 89.3%
- of pages came back dead
Every dot is one page of italyanmutfagi.com, measured 26 July 2026. The bright ones are the pages that get clicked.
Shipping pages is easy. Being seen is not.
People building with AI ship pages fast. The problem is not production, it is that nobody knows whether what was produced actually works.
Measured on our own sites · 26 July 2026
5,442 pages appear in Google; 4,861 bring in not a single click.
source
italyanmutfagi.com, 26 July 2026, last 90 days. Source is the Search Console API at page granularity; 'no clicks' means the clicks field was zero in that window. The page_yield endpoint does the same sum, but this number came straight from the API, not from the endpoint.
A third-party tool sees 248 pages. Google shows 5,442.
source
Same site, same day. The 248 was read from a third-party SEO tool's own interface; 5,442 is Search Console's own figure. They measure different things: what that crawler found, versus what Google actually shows.
15 pages published; Google had not even discovered 7 of them.
source
e.pet, 15 pages published in July 2026. The number comes from the URL Inspection API: for 7 pages Google returned 'URL is unknown to Google', meaning not even crawled. The index_coverage endpoint reads exactly this.
Seven endpoints. Three are free forever.
Everything that comes from your own Search Console costs us nothing, so we do not meter it. You pay only when we have to buy data about someone else.
index_coverageHas Google seen my pagespage_yieldWhich page is dead, why, and which one firstquery_matchWhich one is a step away, which one was never mine
These three never cost you anything, because they never cost us anything. That is exactly what the free plan is: every question about your own site, unmetered. No card required.
competitor_gapWhat do they rank for that I do notkeyword_volumeIs this keyword worth it, for this sitebacklink_profileWho links to me and does it mean anythingserp_positionWho is above me on this query
The schema comes from the live MCP server, not written by hand.
Pricing
Analysing your own site costs us nothing, because the source is Google's free API. That is why we give it away unmetered even on the free plan: it does not count against a quota and it is never billed. An abuse guard caps it at 1,000 calls per site per day; real use does not come close. You only pay when you look at a competitor. If you run many sites the real win is this: instead of opening a dashboard per client, your agent asks all of them and brings back only the ones that need you.
- 1 site
- Your own site: unmetered analysis
- No competitor lookups
- API key + MCP
- Up to 10 sites
- Your own sites: unmetered analysis
- $10 a month of competitor data
- At least 50 lookups (100+ on a small site)
- API key + MCP
- Up to 50 sites
- Your own sites: unmetered analysis
- $20 a month of competitor data
- At least 100 lookups (200+ on a small site)
- Team keys
Two months free on annual billing. What is enforced is that budget, not a lookup count; when it runs out the API says so plainly instead of cutting you off in silence. Work out where in the range you land below.
Which end are you on? Drag it.
What drives the cost of one lookup is how many keywords your own site ranks for. Calculated with the provider prices we measured ourselves.
- $50/mo51 lookups
- $100/mo102 lookups
If you do not know: you can get this number from your own site on the free plan, since Search Console data costs us nothing and that side is unmetered. These are what we PAY the provider, not what we bill you; you are looking at our margin. And the enforced limit is the budget, not a lookup count: when it runs out the API says so plainly. The figures come from the same file the endpoints actually use, not a copy written for this page.
How long it takes to answer
A human waits out a slow page; an agent times out and its loop breaks. So we do not hide the numbers. None of our competitors publish this table.
serp_position132/33611/15keyword_volume624/81218/23backlink_profile64/6511/12competitor_gap892/151626/38Our own benchmark, not customer traffic. Each cell says how many samples it came from: a p95 from five measurements is not the same thing as a p95 from five thousand. As real traffic accumulates this table follows it. Measured: 2026-07-27
Measure it yourself if you like
The numbers above are our own benchmark. The button below sends ten calls in a row from your browser to our live server and works out the round trip on your own network.
Four real sites: what it says today, what happened next
page_yield
The four sites below are ones we run ourselves: italyanmutfagi.com, memuratamalari.com, diolivo.com.tr, business.wiki. Every number is recomputed with the page_yield endpoint's own logic against their live Search Console data: both the bucketing and the main-problem choice are copied from the endpoint's code, none of it hand picked. Switch sites and watch the prescription change.
- 45.6%
- no demand
- 17.8%
- title not earning clicks
- 15.7%
- ranking too low
- 20.8%
- healthy
45.6% of 2,497 pages see no demand, but 20.8% are healthy, the highest of the four sites. There is a pattern that works here; clearing the dead tail and repeating that pattern is the fastest move.
Real measurement · 26 July 2026 · last 90 days · one call, 0.4 seconds, $0
What happened to those same four sites over six months
FUTIA did not produce these gains. They are the result of SEO work on sites we run; FUTIA is the measurement tool we used while doing it, and it came out of that work. The figures were pulled from the Search Console API, not typed by hand. Measured: 2026-07-27 · periods compared: 2026-01-25 – 2026-04-24 / 2026-04-25 – 2026-07-24
The curves are 52 weeks of real weekly clicks; hover one to read any single week. The vertical axis is square-root scaled so one spike week does not flatten the other 51.
No setup, connect in three lines
No setup, no mandatory SDK. Hand the key to your agent and it takes it from there. If you use MCP, one config block is enough.
curl -s https://futia.io/api/v1/page_yield \
-H "Authorization: Bearer $FUTIA_KEY" \
-H "Content-Type: application/json" \
-d '{"site":"sc-domain:example.com"}'Try it right now if you like
The first thing an agent does after connecting to an MCP server is ask for the tool list. The button below sends exactly that request: a real JSON-RPC call from your browser to our live server. No credentials needed.
Your agent already speaks this
MCP is a standard protocol. Our server passes its conformance checks, so none of the clients below need a plugin from you: you put the address in the config and the tools show up. If you would rather not depend on any of them, the same seven endpoints are plain REST.
- Claude Code
- Cursor
- GitHub Copilot
- Windsurf
- Zed
- Cline
- Gemini CLI
- JetBrains
- Warp
- n8n
- LangChain
- ChatGPT
These are compatibility marks, not partnership marks: we have no commercial relationship with any of them and none of them endorse us. The only thing we share is the protocol. Trademarks belong to their owners. ChatGPT appears without a mark because its owner had it removed from the open source icon set we use, and drawing one from memory would be worse than writing the name.
Side by side with Semrush
Their 84 reports are a strength, not a gap. But that strength costs a decision: before the call you need to know which report you want, which unit it bills in, and how many rows it will return. Pick a toolbox and see the load for yourself.
- reports
- 15
- unit range
- 40 · 100
- satır
- istek
You cannot know the unit count BEFORE the call: nobody knows how many rows come back until it returns.
- What this call costs us
- 0.6650 $
A known number beats an unknown invoice. This is what we pay the provider; what you pay is the monthly plan.
Every box on the left was measured, not typed: the report counts and price ranges come from ten discovery calls to Semrush's OWN MCP server, and the unit values are their `price` field itself. 28 July 2026. One toolkit was closed on our subscription and is not counted, so 84 is a lower bound. The plan price is a separate source, the published pricing page: API access sits on the Advanced plan at $455.67/mo billed annually. Prices change, so we state the date.
What happens to your data
We hold no certifications and we are not going to claim any; this product has not launched. What we can tell you is what we actually do:
- We request read-only access. We make no changes to your site and hold no write permission.
- Your Google tokens are stored encrypted, and the encryption is bound to your account: a copied database row cannot be decrypted under another account.
- We never store your API key itself, only a digest of it. That is why it is shown once.
- Your Search Console data is never written to disk. It is read, turned into the answer, and gone when the request ends. There is no table for it.
- What we do cache is what we buy from the data provider, for 24 hours. That is public data about a domain, not yours.
- You can revoke access from your Google account in one click. You do not need our permission: myaccount.google.com/permissions
- Your data is never sold and never used for advertising. The competitor endpoints do pass the domain and keyword you asked about to our data provider, because that is how the answer is bought; nothing that identifies you goes with it.
Questions, answered honestly
Really no dashboard?
None. The product is an API and an MCP server. You hand your agent the key, it asks, it gets the answer. There is no screen to look at.
Do I have to connect Search Console?
Yes. Your site's real search data exists only there. We measured it: third-party databases see only 4% to 10% of a site's search presence. Without the connection, anything we told you would be a guess.
What happens to my data?
The one sentence: your Search Console data is never written to disk. All seven commitments live in the "What happens to your data" section above, in one place, so the two cannot drift apart.
My site is new with no traffic. Is this useful?
Yes. The index_coverage endpoint needs no traffic; it tells you whether Google has seen your pages at all. The other two gain meaning as traffic arrives.
Give your agent a key and let it handle the rest
Connect your Google account, create a key, hand it to your agent. No card required.
Connect Google, start free