WebScore MCP Server: Audit Websites From Claude, Cursor & Codex

Most website audits break your flow. You're deep in your editor, you notice a page feels slow, and now you have to stop, open a browser, paste the URL into a tool, wait, read a report, and translate its findings back into code. The WebScore MCP server removes that round trip: your AI assistant runs the audit and reads the results for you, right where you're already working.

This guide covers what the MCP server does, the tools it exposes, how to connect it in a couple of minutes, and the use cases where it genuinely changes how you work.

What is the WebScore MCP server?

The Model Context Protocol (MCP) is an open standard that lets AI assistants call external tools. WebScore ships a hosted MCP server, so any MCP-capable client — Claude Desktop, Claude Code, Cursor, Codex, Grok, and others — can drive a full WebScore audit through plain language.

Instead of switching to a browser, you ask your assistant to "scan example.com with WebScore" and it comes back with per-module scores, the specific issues, how to fix each one, and domain authority — as structured data it can act on.

The key difference from a normal audit tool: your assistant receives the fixes, not just the numbers. So it can go straight from "here's what's wrong" to "here's the patch."

The tools you get

Connecting the server adds ten tools to your assistant:

Why it's better than tab-hopping

• No context switching. The audit happens in the same window as your code, your terminal, and your AI.
• Actionable, not just scores. scan_website returns each issue with a concrete fix, so your assistant can apply it — not just report it.
• It coordinates with your account. Scan a new site through MCP and it's added to your WebScore dashboard automatically, so the results show up there too.
• One key, many clients. The same personal key works across Claude Desktop, Claude Code, Cursor, and any MCP client.

Real use cases

1. Fix a performance score end-to-end

The single biggest win. Instead of reading a Lighthouse report and hand-translating it, you let the assistant close the loop:

Scan mysite.com with WebScore, then fix the top 3 performance
issues you find in this repo.

Your assistant calls scan_website, reads issues like "Reduce unused JavaScript — code-split and lazy-load below-the-fold bundles," locates the relevant files, and proposes the change. Audit and fix in one session.

2. A pre-deploy gate in Cursor

Wire the audit into your review ritual. Before you merge, ask:

Scan the staging URL and tell me if Security or Accessibility
dropped below 90. If so, list what regressed.

Now a score regression is something you catch at review time, not from an angry customer.

3. Email deliverability triage

DNS problems are invisible until mail stops landing. From the chat:

Check SPF, DKIM and DMARC for acme.com and tell me what's missing.

check_email_auth reports exactly which records are configured, so you can fix deliverability before a campaign goes out — no dashboard required.

4. Competitive and domain research

scan_website includes domain metrics (Domain Rating, Domain Authority, Trust/Citation Flow, backlinks and referring domains):

Scan competitor.com — what's its domain rating and backlink
profile compared to mine?

Great for sizing up a niche or reporting to a client without opening five SEO tools.

5. Bulk site triage

If you manage several sites, let the assistant find the weak spots:

List my WebScore sites and tell me which three have the lowest
overall score, then scan the worst one and summarize the fixes.

list_websites surfaces the portfolio; scan_website drills in.

6. Turn on monitoring by voice

Enable weekly monitoring for mysite.com.

set_monitoring flips it on, and WebScore re-scans on schedule and emails you when scores drop.

Connect it in two minutes

1. Open Settings → API & MCP and copy your key (MCP access is a Lifetime perk).
2. Add WebScore to your client's MCP config:

{
  "mcpServers": {
    "webscore": {
      "url": "https://webscore.now/api/mcp",
      "headers": { "Authorization": "Bearer <your-key>" }
    }
  }
}

3. Restart the client and ask: "scan example.com with WebScore."

The server uses a simple Bearer API key, which works today in Claude Desktop, Claude Code, Cursor and most MCP clients that let you set an auth header. Full setup details live in the MCP docs.

Security and access

• Your key is personal and revocable — regenerate it any time in Settings, which instantly invalidates the old one.
• Access is scoped to your account. A scan you run is tied to your projects, and new sites are added to your dashboard so nothing is hidden away in the chat.
• MCP access is part of Lifetime — one payment, then unlimited scans from your AI.

The bottom line

The best audit is the one you actually run. By putting WebScore inside the tool you already build with, the MCP server turns "I should check that site's performance/SEO/security" from a task you postpone into a sentence you type. Your AI scans, reads the fixes, and — if you want — writes them.

Get Lifetime access to connect the WebScore MCP server, or read the setup docs first.