claude code
❯
How it works
The loop you stop thinking about.
The agent watches, Search Console data grades every fix, Claude ships them.
The agent watches
Weekly auto-scans find what changed since your last deploy. New issues, regressions, and opportunities caught without you asking.
Search Console grades the fix
Your Search Console metrics sync daily per URL. After a fix ships, the agent tracks clicks, impressions, and position against the baseline. No generic severity scores, real traffic decides whether it worked.
Claude ships the fix
Connect via MCP. Claude Code, Cursor, or Windsurf reads the issue and writes the fix in your codebase. You approve, it commits.
Slow Largest Contentful Paint (4.0s)
LCP is exactly at Google's “poor” threshold of 4s. One more ms and Core Web Vitals fail.
Why this matters here
Your hero image is the LCP element, and blocking scripts in <head> delay it by ~1s. Compressing and prioritizing the hero is the fastest ranking win.
Fix prompt
In app/layout.tsx, add fetchPriority="high" to the hero <img>, serve it as WebP under 80KB, and add defer to non-critical scripts.
Step 1 · Scanner
The agent watches your site.
Every issue comes back as broken HTML plus an AI-ready fix prompt. Run scans weekly on autopilot, or on demand from MCP, CLI, or dashboard.
- Broken element extracted from your page, not a generic warning
- Fix prompt designed for Claude Code, Cursor, or any AI agent
- 40+ checks across SEO, performance, accessibility, and AEO
Step 2 · Memory
Memory remembers every scan.
The agent carries full context to every conversation. Ask Claude what is broken or what to build next, it already knows the site history.
- Issues labeled NEW, PERSISTING, REGRESSED across scans
- Site profile: goal, audience, niche inferred on first scan
- Visual review: screenshot scored against your ICP
- Content gaps: missing pages suggested from your sitemap
seolint.dev
7 scans · 42 days tracked · 50 URLs in sitemap
Slow LCP (4.0s) on homepage
2h agoMissing og:image on /pricing
3d agoH1 missing on /blog
5d agoContent gaps · 3 page ideas
Getting started with SEOLint
/docs/quickstart
SEOLint vs Ahrefs: SEO for AI agents
/vs/ahrefs
Insight: Strong blog coverage but no comparison pages vs direct competitors. One-hour content win.
❯ Is my site improving? Fix anything critical.
● get_site_status (MCP)
⎿ Improving · 6 scans · 32d tracked
● scan_website (MCP)
⎿ 1 new · 1 persisting
● Edit(app/layout.tsx)
- <title>Home</title>
+ <title>Best CRM for small teams</title>
1 critical issue fixed. Memory updated.
Step 3 · MCP
Fix it inside Claude.
Connect SEOLint as an MCP server. Claude Code, Cursor, or any AI agent reads the issue and writes the fix in your codebase. You approve, it commits.
- Works in Claude Code, Cursor, Windsurf, Claude Desktop, and custom agents
- 8 tools: scan, history, site intelligence, page suggestions, resolve, and more
- GitHub Actions: fail builds when fixed issues regress
Step 4 · PR agent
Coming soonAnd the agent writes the PR too.
Connect a GitHub repo and the agent opens a pull request with the fix, waits for your CI to go green, then hands it to you for review. Mechanical fixes first: meta, alt, canonical, schema, sitemap.
acmeshop/site · #142
fix(seo): add meta descriptions to 3 product pages
Flagged in last 3 scans. Drafted descriptions based on page content and your existing voice. Each is 140–155 chars.
+ app/products/trail-shoes/page.tsx
+ app/products/waterproof/page.tsx
+ app/products/lightweight/page.tsx
Try it free for 7 days.
Card required, no charge until day 8. Everything included, one flat price per site.
SEOLint
$199$99/mo
1 site · agent + MCP · everything included
7-day free trial, no charge until day 8
Start 7-day free trialStop thinking about SEO.
Get started for freeListed on
