A CrUX MCP: Chrome UX Report field data, as a tool your AI assistant can call
Ask how your site performs for real visitors and get Google's own field data back, not a guess. The GSC Wizard MCP gives Claude, ChatGPT and any MCP-capable assistant nine months of weekly LCP, INP and CLS, plus FCP and TTFB, for the whole site or a single page, with ratings and regressions already worked out.
Same connector endpoint: mcp.gscwizard.com
Field data, not a lab test
LCP + INP + CLS, week by week
The tool reads the Chrome UX Report History API: the 75th percentile of what real Chrome users experienced on your site, the same dataset Search Console's Core Web Vitals report is built on. One call returns up to 40 weekly data points, each covering a rolling 28 days.
It lives on the same MCP as your Search Console, GA4 and Bing tools, so the assistant can set a ranking drop next to the weeks your page experience changed. The model reads the numbers, it does not invent them.
One read tool, get_core_web_vitals, covers all of it. Ask in plain language; the assistant picks the device and the target.
Largest Contentful Paint, Interaction to Next Paint and Cumulative Layout Shift: the 75th-percentile value for every week, the latest rating, and the share of visits in the good bucket over time.
First Contentful Paint and Time to First Byte come back in the same call and are rated the same way. They are not Core Web Vitals, but a slow server shows up in TTFB before it shows up in LCP.
Every response lists the Core Web Vitals that dropped a rating since the previous week, for example good to needs improvement, with both values and both weeks, so the assistant does not have to eyeball a series.
All devices together, or one device class at a time. Mobile and desktop often tell different stories, and Google evaluates them separately.
By default the tool measures the property's origin. A domain property maps to its https origin, and when CrUX has no record for it the tool tries the www (or non-www) twin and says which one it measured. Pass a URL to measure a single page.
Where Chrome detected ads, the response adds CrUX's four experimental ad metrics: ad count, ad density, ad CPU and ad network weight, each as a weekly trend. Chrome sets no targets for them, so they are reported, never rated.
Each metric comes back with its 75th-percentile value, a formatted value and a rating against Google's published Core Web Vitals thresholds. The assistant reports "good" because the response says good, not because it remembered a number.
| Metric | Good | Poor | Core Web Vital |
|---|---|---|---|
| LCP | ≤ 2.5 s | ≥ 4 s | Yes |
| INP | ≤ 200 ms | ≥ 500 ms | Yes |
| CLS | ≤ 0.1 | ≥ 0.25 | Yes |
| FCP | ≤ 1.8 s | ≥ 3 s | No |
| TTFB | ≤ 800 ms | ≥ 1.8 s | No |
Anything between the two columns is "needs improvement". Regressions are only reported for the three Core Web Vitals.
{
"siteUrl": "sc-domain:gscwizard.com",
"target": { "origin": "https://www.gscwizard.com" },
"formFactor": "ALL",
"periodCount": 40,
"latestPeriod": "2026-10-03",
"metrics": [
{
"label": "LCP",
"name": "Largest Contentful Paint",
"coreWebVital": true,
"latestP75": 1403,
"latestP75Formatted": "1.40 s",
"latestRating": "good",
"latestGoodDensity": 0.9249,
"p75Series": [ ..., 1442, 1404, 1451, 1403 ]
},
{
"label": "INP",
"latestP75Formatted": "49 ms",
"latestRating": "good",
...
},
...
],
"regressions": []
}The response for this website on 10 October 2026. The property is a domain property; CrUX has no record for the bare domain, so the tool measured the www origin and says so in target.
CrUX is the closest thing to how Google measures page experience, but it has limits of its own. The tool is built so the assistant reports them instead of papering over them.
noData with a link to Chrome's coverage docs, not an error and not a made-up value.Type these into Claude, ChatGPT or any MCP client in your own words. It reads from the same properties you already have in GSC Wizard.
"How have my Core Web Vitals trended on mobile over the last nine months?"
"Did any Core Web Vital drop a rating this week? Which one, and from what to what?"
"Compare LCP on phone and desktop and tell me how big the gap is."
"Check INP for our checkout page on its own. Is it worse than the site as a whole?"
"We switched ad partners last month. Did ad CPU or ad density go up, and did INP move with it?"
"Mobile clicks fell in September. Did our Core Web Vitals on phones change in the same weeks?"
In Google Cloud, enable the Chrome UX Report API and create an API key for it. Google does not charge for the CrUX API. Chrome's CrUX API docs walk through it.
Open a property and pick Core Web Vitals in the left sidebar, under Analytics. Use the API Key button on that report. The key is stored encrypted, and the same key runs the in-app report and its weekly regression alerts.
Add mcp.gscwizard.com as a connector in your AI assistant (or paste an API key). The tool only reads, so a read-only key or a read-only connector approval is enough. If no CrUX key is saved yet, it answers with a short notConfigured note that names these steps instead of an error.
Works with Claude, ChatGPT, Gemini, Grok and any MCP-capable client; see the configs for every MCP client. Prefer charts? The same data is a report in the app, with weekly email, Slack and Teams alerts: see Core Web Vitals tracking.
An MCP (Model Context Protocol) server is a standard connector that lets an AI assistant such as Claude or ChatGPT read live data and call tools. A Core Web Vitals MCP, also called a CrUX MCP, gives the assistant your Chrome UX Report field data, so it can read how real Chrome users experienced your site's loading, responsiveness and layout stability instead of guessing or running a lab test.
From Google's Chrome UX Report (CrUX) History API: field data collected from real Chrome users, the same dataset Search Console's Core Web Vitals report is built on. Each data point is the 75th percentile over a rolling 28-day window, published weekly, and the tool returns up to 40 of them, about nine months of history.
Chrome only publishes field data for origins and pages with enough real Chrome traffic, so small sites and individual pages are often missing. The tool then answers noData with a link to Chrome's coverage docs instead of an error. Try the whole site instead of one page. For a whole-site request the tool already tries both the www and the non-www origin.
No. Lighthouse runs a lab test of one page load on a simulated device. The Core Web Vitals MCP reads field data: what real Chrome visitors experienced over the past 28 days, week by week. It runs no lab test; use PageSpeed Insights or Lighthouse for that.
It reads CrUX with a Chrome UX Report API key, which is free from Google Cloud. Save it once on the Core Web Vitals report in the GSC Wizard app, with the API Key button. The same key runs the report and its weekly regression alerts. Without a key the tool answers notConfigured, with the steps to add one.
MCP access is included in every GSC Wizard plan, and every new account starts with a 7-day free trial. The Chrome UX Report API itself is free. See pricing.
The core of the server: clicks, impressions, rankings, decay and indexing, read from a warehouse of your full history.
Tell Bing, Yandex and the other IndexNow engines that pages changed, with the key file checked before anything is sent.
Sessions, key events, ecommerce revenue and AI / LLM traffic, blended with Search Console.
The same CrUX history as charts in the app, with weekly regression alerts by email, Slack or Teams.
Nine months of CrUX field data for LCP, INP and CLS, by device, with ratings and regressions worked out, on the same connector as Search Console.
Practical GSC and GA4 workflows, product updates and SEO experiments. No spam, unsubscribe anytime.