MCP connector · PageSpeed Insights

Page speed & Core Web Vitals, answered by your AI assistant

Ask “why is this page slow?”, “do we pass Core Web Vitals?” or “are we faster than our competitors?” - the assistant runs Google PageSpeed Insights and reads real-user Chrome data, then tells you exactly which files to fix. Nothing to connect: any public URL works.

Sign in / Sign up with GoogleSign in / Sign up

We couldn't sign you in

If this keeps happening, our support team will help - create a ticket or contact us.

Also connect Google Search Console
  • No Google account to connect
  • Free monthly credits
  • Pay only for successful calls
Example conversation
Why is our pricing page slow on mobile? What should we fix first?
pagespeed_get_opportunities2 credits

Real users wait 4.1 s for the main content (LCP, poor). The three biggest fixes:

FixSaves
Defer /assets/app.css + 2 fonts1.2 s
Resize hero.png (1.4 MB → WebP)0.9 s
Remove unused chat-widget.js0.4 s

The hero image is also the LCP element - preload it and serve it at 1200 px. That alone should bring LCP under 2.5 s.

Works withClaudeChatGPTClaude CodeCursorVS CodeAny MCP client

Setup

Connect your AI app

Sign in, then follow the steps for your app on the setup page.

Server URLhttps://toolkitbackend-j69m.onrender.com/api/v1/pagespeed/mcp
  1. Click "Open Claude connectors" below - claude.ai opens on Settings → Connectors.

  2. Click "Add custom connector". Name: PageSpeed Insights. URL: use the Copy button above and paste it.

  3. Click Add, then Connect. Sign in to 500Toolbox and click Allow. Done.

  4. In a chat, open the tools menu (the + / sliders icon) and switch PageSpeed Insights on.

A connector added on claude.ai also appears in the Claude desktop and mobile apps. On Team and Enterprise plans an owner adds it once under Organization settings → Connectors; members then click Connect.

Why ask your assistant

From a score to a to-do list

The exact files to fix

Not just “reduce unused JavaScript” - the scripts, images, stylesheets and elements responsible, ordered by the time each fix saves.

What real users experience

Chrome UX Report data for your page or whole site - the 28-day numbers Google uses for ranking - and up to 40 weeks of history.

You vs your competitors

Real-user Core Web Vitals for up to 5 sites side by side, ranked, with where each one is weakest.

Core Web Vitals

The three numbers Google ranks on

A page passes when 75% of real visits meet all three. Every tool rates each metric against these thresholds.

LCPLargest Contentful Paint

How long until the main content shows.

Good ≤ 2.5 sPoor > 4 s
INPInteraction to Next Paint

How quickly the page reacts to taps and clicks.

Good ≤ 200 msPoor > 500 ms
CLSCumulative Layout Shift

How much the layout jumps while loading.

Good ≤ 0.1Poor > 0.25

How it works

Set up once in about a minute

1

Sign in to 500Toolbox

That is all the setup there is - PageSpeed and Chrome UX Report data are public, so there is no Google account to connect.

2

Add one URL to your AI app

Paste the server URL as a custom connector in Claude, ChatGPT, Cursor or VS Code and click Connect.

3

Ask about any page

Yours or a competitor's. The assistant picks the right test, runs it and explains the result - every number comes from Google.

What you can ask

Questions it answers with Google's speed data

“Why is my homepage slow on mobile, and what should my developer fix first?”

“Do my product pages pass Core Web Vitals for real users?”

“Did last month's speed fix actually improve LCP?”

“How does my site's speed compare to my top 3 competitors?”

“Which of these 10 landing pages is the slowest?”

“Which images and scripts are making this page heavy?”

9 tools

Everything your assistant can test

Lighthouse tests that say what to fix, and Chrome UX Report data that says whether real users are affected. Every tool name starts with pagespeed_, so it never gets mixed up with your other connectors.

Lighthouse tests (what to fix)

Audit a page2 cr

Performance, accessibility, best-practices and SEO scores, lab metrics, real-user vitals and the top issues.

pagespeed_audit_page
Fixes with the files to change2 cr

Every failing speed audit by time saved, with the exact scripts, images, CSS and elements responsible.

pagespeed_get_opportunities
Mobile vs desktop3 cr

Scores, metrics and real-user data side by side, plus issues that only hurt mobile.

pagespeed_compare_devices
Audit up to 10 pages1 cr per page

One row per page: score, LCP, TBT, CLS, Core Web Vitals pass/fail and the biggest fix.

pagespeed_audit_pages

Real-user Core Web Vitals (what Google ranks on)

Real-user Core Web Vitals1 cr

LCP, INP, CLS, FCP and TTFB from real Chrome users over 28 days, rated against Google's thresholds.

pagespeed_get_field_data
Core Web Vitals trend1 cr

Up to 40 weeks of real-user vitals - see whether a fix worked or when a regression started.

pagespeed_get_vitals_history
Check up to 20 pages2 cr

Pass / fail on Core Web Vitals for many pages at once, with the failing metric for each.

pagespeed_check_core_web_vitals
Compare with competitors2 cr

Real-user speed of up to 5 sites, ranked, with who passes and where each is weakest.

pagespeed_compare_sites

Account

Account statusFree

Credit balance and the price of every tool.

pagespeed_get_account_status

One-click reports

Built-in prompts

Apps that support MCP prompts (Claude shows them in the + menu) get complete speed workflows - no need to know what to ask.

Page speed report

Scores, real-user Core Web Vitals, the trend and a prioritised fix list.

Fix a slow page

A developer to-do list grouped by LCP, interactivity, layout shift and page weight.

Compare speed with competitors

Your site against competitors on real-user data, and how to close the gap.

Core Web Vitals audit

Your key pages checked, failing templates found, one fix per template.

Security & privacy

Nothing private involved

Public data only

Tests fetch the public page, the way any visitor would. No logins, cookies or private pages are involved.

You approve every app

An AI app gets access only after you click Allow on a 500Toolbox page that shows which app is asking.

Read-only

Nothing on your site is changed. The tools only measure and report.

Remove apps anytime

Disconnect any AI app from your dashboard in one click.

Pricing

Pay per test, from 1 credit

Free credits every month, plus a welcome grant when you sign up. Failed tests are refunded automatically, and purchased credits never expire.

1 crreal-user data, per page in bulk
2 crfull audit, fix list, checks
3 crmobile vs desktop
Credit packs →

FAQ

Questions people ask

Do I need to connect anything?

No. PageSpeed Insights and the Chrome UX Report are public Google data, so any public URL can be tested - yours or a competitor's. Sign in to 500Toolbox, add the server URL to your AI app and ask.

What is the difference between lab and field data?

Lab data is one Lighthouse test run now on Google's servers, with a simulated mid-range phone on slow 4G for mobile. It is repeatable and tells you what to fix. Field data is what real Chrome users experienced over the last 28 days (the Chrome UX Report) - that is what Google uses for the page experience ranking signal. The assistant uses both: field data to judge, lab data to diagnose.

When does a page pass Core Web Vitals?

When, for 75% of real visits, Largest Contentful Paint is 2.5 s or less, Interaction to Next Paint is 200 ms or less and Cumulative Layout Shift is 0.1 or less. These are Google's published thresholds, and every tool rates each metric as good, needs improvement or poor against them.

Why does my page have no real-user data?

Google only publishes Chrome UX Report data for pages and sites with enough Chrome visits. A quiet page often has none, so the tools fall back to the whole site's figures and say so. A new site may have none at all - then the lab test is your guide.

Why does the score change between runs?

Lighthouse simulates a device and network, and small differences in server response, third-party scripts and Google's machines move the score a few points. Treat a change of 5 points or less as noise; real-user data over 28 days is much steadier.

What does it cost?

Each tool shows its credit price: a full audit or fix list is 2 credits, mobile-vs-desktop 3, real-user data 1, checks of many pages 2, and the bulk audit 1 credit per page. Your account gets free credits every month, and you are only charged when a call succeeds: anything that fails is refunded automatically.

Your page speed, one question away

Add it once - every chat after that can test any page.

Sign in / Sign up with GoogleSign in / Sign up

We couldn't sign you in

If this keeps happening, our support team will help - create a ticket or contact us.