02 / 04Case study / 02 of 04

MCP server / SEO

seo-mcp

Engineering + open source

In one line

SEO, in your editor.

A free, open-source MCP server that brings professional SEO auditing into Claude Desktop, Claude Code and Cursor.

The seo-mcp website: a “Free & open-source MCP server” badge above the headline “Scan any website for SEO issues”, a line on its 28 SEO rules, site crawling, deep page analysis and monitoring for Claude Desktop, Claude Code and Cursor, and the Get Started and View on GitHub buttons

Year

2026

Platform

MCP server

Stack

All technologies

TypeScript
MCP servers
Node.js
Worker Threads
Claude AI

Also: Node.js · Worker Threads · Claude AI

Live seo-mcp.dev

01 / The problem

SEO, siloed.

Professional SEO tools are expensive, need API keys and live outside your development workflow.

Developers and marketers switch between multiple tabs, wait for reports to generate and can’t ask follow-up questions in natural language.

SEO stays siloed from the AI-powered tools that now power everything else.

Expensive tools
API keys required
No follow-up questions

02 / The solution

Just ask.

An MCP server that plugs professional SEO auditing straight into your AI editor.

seo-mcp is a free, open-source Model Context Protocol (MCP) server for AI editors like Claude Desktop, Claude Code and Cursor. Ask your AI assistant to audit a page, crawl a site or monitor content changes, and it does, without leaving your editor.

No API keys. No subscriptions. Just run it locally.

How it flows

Runs on your machine

Local

01 / Ask

You ask your AI editor to audit a page, crawl a site or watch for changes

02 / Crawl

Node.js worker threads scan pages in parallel

03 / Audit

28 rules return structured, severity-ranked findings

04 / Answer

Your AI assistant explains and prioritizes the fixes

03 / Key features

Four things it does well.

Every result is structured for an AI assistant, so it can explain and prioritize fixes, not just find issues.

F.01

Severity-ranked

Meta tags
Content
Technical
Performance

seo-mcp · scan_url
url: https://larch.ee/work/seo-mcp/

Score: 90/100
Status: 200 | Response Time: 206ms | Page Size: 64.1 KB
Word Count: 609
Issues Found (2)
- [meta-title-length] Title length is 64 characters (recommended: 30-60)
- [meta-description-length] Meta description is 188 characters (recommended: 120-160)
Images: 2 total (0 missing alt)

28 SEO audit rules

Meta tags, content quality, technical factors and performance signals. Every rule returns structured, severity-ranked findings optimized for LLM interpretation.

rules: 28

F.02

Configurable depth
Page limits

seo-mcp · scan_site
url: https://larch.ee/  maxPages: 40

Pages Scanned: 26
Total Issues: 32
Scan Duration: 5.6s
Issues by Category
  Meta: 28
  Content: 4
Robots.txt: Found
Sitemap: Found (26 URLs)

Full site crawling

Point it at any domain and it crawls the entire site, with configurable depth and page limits. Node.js worker threads keep scans fast, even on large sites.

F.03

Firehose
SSE
Lucene

Real-time monitoring

Live website monitoring through the Firehose (Ahrefs) service. Content changes stream in as they happen over SSE, filtered with Lucene query syntax.

F.04

DOM
Link graphs
Resources

seo-mcp · analyze_page
url: https://larch.ee/work/seo-mcp/

Score: 90/100
Status: 200
Response Time: 128ms
Page Size: 64.1 KB
H1: seo-mcp
Word Count: 609
Internal Links: 36
External Links: 12
Images: 2 (0 missing alt)

Deep page analysis

Rich structured data for every page: full DOM inspection, link graphs, resource inventory and issue breakdowns, everything an AI agent needs to reason about your SEO health.

04 / Built with

How it’s built.

Server
- Node.js
- TypeScript
AI editors
- MCP
- Claude

05 / Results

The outcome

seo-mcp / 2026

Teams get professional-grade SEO auditing inside their existing workflow, all for free.

They can ask follow-up questions, run targeted fixes and monitor changes, all in natural language. The code is open, so anyone can inspect the ruleset and contribute new audit rules.

Visit seo-mcp.dev