The MCP server for website monitoring
Ask Claude or ChatGPT how your website is really doing, and get an answer from your own monitoring data. Whether anything is down, how fast your pages are for real visitors, what Google is showing people, and what changed this week.
Which pages regressed on LCP this month?
Three pages dropped out of the good range since June 1. The biggest mover is /pricing.
Everything you monitor, answerable in a sentence
It can reach your Core Web Vitals, your uptime, your search traffic, your indexing and your robots.txt: 50+ tools in all. Most of them read, and some act on your behalf when you ask. Your assistant picks the right one and hands you the answer.
Core Web Vitals
Field LCP, INP and CLS, and how they have moved over time.
Real User Monitoring
What real visitors experience page by page, plus the JavaScript errors they hit.
Uptime and SSL
Whether the site is up, what went down and when, and which certificates expire soon.
Synthetic monitoring
Your scheduled checks and the results they came back with.
Google Search Console
Clicks, impressions, the queries behind them and the pages they land on.
Indexing
What Google has indexed, and what is stuck outside the index.
Web Analytics
Traffic, sources, audience, events and ecommerce, without opening a dashboard.
Robots.txt and sitemaps
The current file, what changed in it, and whether a given URL is allowed.
It can also change things when you ask it to: add a CrUX URL, edit an alert rule, generate a report. Nothing it does exceeds what your own role in the workspace allows. See every tool in the docs.
It is a Google Search Console MCP server too
Your Search Console data is already in VitalSentinel, so it is already here. Ask about clicks and impressions, the queries behind them and the pages they land on, what Google has indexed and what is stuck outside the index, and what your robots.txt and sitemaps are doing about it. Connect Search Console to VitalSentinel once, and every assistant you connect afterwards can read it – no second connector, nothing extra to authorize.
The difference from a Search Console connector on its own: the answer arrives beside your Core Web Vitals, your uptime and your analytics, so “did that ranking drop line up with the LCP regression?” is one question rather than three exports.
Your assistant stops guessing about your site
An assistant that cannot see your site can only give you general advice. Connect it and the same questions come back answered from your own measurements: what your visitors actually experienced, on your pages, over the period you asked about.
- Answers built from your live data, not from generic best practice
- One question can span search, performance and uptime at once
- Follow-up questions keep the same context, so "why?" works
Connect it once, then forget it is there
MCP is simply the standard way an AI assistant connects to a service, and you do not need to know anything more about it than that. Adding VitalSentinel works like linking any other account: nothing to install, and nothing to keep running afterwards.
Add the address
Paste the server address into your assistant’s connector settings. There is no key to generate first.
Approve it in your browser
Sign in the way you always do, choose which workspaces it may reach, and approve.
Just ask
From then on it answers from your live data, and you can revoke it whenever you like.
The address you paste in step one:
https://mcp.vitalsentinel.com/mcp
In Claude Code, all three steps are a single command:
claude mcp add --transport http vitalsentinel https://mcp.vitalsentinel.com/mcp
Works in
- Claude Code
- Claude
- ChatGPT
- Cursor
- VS Code
- Notion
The same address works in all of them, and you can disconnect any of them whenever you like.
Included on every plan, Free included. Nothing to buy, nothing to upgrade, no waiting list.
The numbers are current, not last Tuesday's
This is not a copy of your data sitting in a file somewhere. Every question reads the same live data the app shows, so what your assistant tells you is what your dashboard would show if you opened it.
- The same data the app shows, read at the moment you ask
- Fix something today and tomorrow the answer moves
- Nothing to keep in sync, because nothing is copied
It holds no credentials of its own
No database access, no service account, no shared key sitting somewhere waiting to leak. Connecting is an ordinary sign-in in your own browser: the server carries your own sign-in and stores nothing.
- You choose which workspaces it can reach
- Every permission it asks for is listed up front
- Never more access than your own role allows
- Revoke the connection whenever you want
Permissions are checked against your current role on every single request.
And if you would rather build on the data directly
There is a full REST API over the same monitoring data, on every plan including Free, for anyone putting together their own dashboard or report. MCP and API usage draw on the one workspace credit allowance.
https://api.vitalsentinel.com/api/public/v1
Give your assistant something real to work with
Create a free account, add a domain, and connect it to Claude or ChatGPT in about a minute. Your monitoring data becomes something you can simply ask about.