SEO Check MCP
A Model Context Protocol (MCP) server that provides comprehensive SEO analysis tools with actionable fix instructions for AI assistants like Claude Code and Claude Desktop.
README
SEO Check MCP
A Model Context Protocol (MCP) server that provides comprehensive SEO analysis tools with actionable fix instructions for AI assistants like Claude Code and Claude Desktop.
Features
- 24 SEO Analysis & Fix Tools
- Actionable Instructions for Claude Code to apply fixes directly
- Dual Transport Support: HTTP (Claude Code) and stdio (Claude Desktop)
- Headless Chrome via Selenium for screenshots and Core Web Vitals
Quick Start
Prerequisites
- Bun runtime
- Chrome/Chromium browser (for screenshot tools)
Installation
git clone git@github.com:muningis/seo-check-mcp.git
cd seo-check-mcp
bun install
Claude Code Integration (HTTP Server)
Claude Code supports HTTP-based MCP servers:
1. Start the server
bun server.mts
2. Configure Claude Code
Add to your project's .mcp.json:
{
"mcpServers": {
"seo-analysis": {
"url": "http://localhost:3000/mcp"
}
}
}
Or add to your global ~/.claude/settings.json:
{
"mcpServers": {
"seo-analysis": {
"url": "http://localhost:3000/mcp"
}
}
}
3. Use the tools
The 24 SEO tools will be available in Claude Code automatically.
Claude Desktop Integration (stdio)
Add to your claude_desktop_config.json:
{
"mcpServers": {
"seo-analysis": {
"command": "bun",
"args": ["cli.mts"],
"cwd": "/path/to/seo-check-mcp"
}
}
}
Replace /path/to/seo-check-mcp with the actual path to this project.
Available Tools (24 total)
Core Tools
| Tool | Description |
|---|---|
read-sitemap |
Fetches and parses XML sitemaps |
read-robots-txt |
Fetches robots.txt content |
scan |
Comprehensive page analysis with optional screenshots |
scan-lite |
Lightweight page scan (no WebDriver) |
Content Analysis
| Tool | Description |
|---|---|
analyze-content |
Keyword density, readability scores, content suggestions |
check-readability |
Flesch-Kincaid, Gunning Fog, SMOG analysis |
Technical SEO
| Tool | Description |
|---|---|
validate-schema |
JSON-LD/Schema.org validation with @graph analysis and completeness scoring |
analyze-sitemap |
Deep sitemap analysis (freshness, priorities) |
check-internal-links |
Internal linking structure analysis |
find-broken-links |
Broken link detection with status codes |
Accessibility & Social
| Tool | Description |
|---|---|
analyze-semantic |
Semantic HTML and ARIA accessibility analysis (WCAG 2.1 Level A) |
analyze-social-meta |
Open Graph, Twitter Cards, and Facebook meta analysis with platform previews |
On-Page Optimization
| Tool | Description |
|---|---|
analyze-meta |
Meta tag analysis with SERP preview |
analyze-headings |
H1-H6 structure validation |
analyze-images |
Image SEO with alt text suggestions |
analyze-url |
URL structure and canonical analysis |
Competitive Analysis
| Tool | Description |
|---|---|
compare-pages |
Side-by-side SEO comparison |
benchmark-seo |
Overall SEO score with grades |
Fix Tools (Actionable Instructions)
These tools return structured instructions that Claude Code can execute:
| Tool | Returns |
|---|---|
fix-meta |
Title, description, canonical, Open Graph fixes with HTML snippets |
fix-images |
Alt text fixes with suggested values and CSS selectors |
fix-headings |
Heading hierarchy fixes with before/after HTML |
fix-schema |
JSON-LD schema templates ready to add |
generate-seo-tasks |
Prioritized task list with recommended tools |
improve-content |
SEO, readability, and structure improvements for local markdown files |
Example Workflow with Claude Code
1. Generate a task list
Use the generate-seo-tasks tool on https://example.com
2. Review prioritized improvements
The tool returns tasks sorted by priority (critical → low) with effort/impact ratings:
{
"url": "https://example.com",
"score": 62,
"tasks": [
{
"id": "meta-title-missing",
"description": "Add missing title tag",
"tool": "fix-meta",
"priority": "critical",
"effort": "low",
"impact": "high"
}
],
"quickWins": ["meta-title-missing", "meta-canonical"],
"summary": "SEO Score: 62/100. 1 critical issue(s) need immediate attention."
}
3. Apply fixes
Run fix-meta on https://example.com to get the HTML changes needed
4. Example fix-meta output
{
"url": "https://example.com",
"instructions": [
{
"action": "replace",
"target": { "type": "html-tag", "selector": "title" },
"value": {
"current": "<title>Home</title>",
"suggested": "<title>Example - Your Brand Name</title>"
},
"reason": "Title too short (4 chars). Should be 50-60 chars with keywords.",
"priority": "critical",
"automated": true
},
{
"action": "add",
"target": { "type": "html-tag", "selector": "head", "tagName": "link" },
"value": {
"suggested": "<link rel=\"canonical\" href=\"https://example.com\">"
},
"reason": "Missing canonical URL causes duplicate content issues.",
"priority": "high",
"automated": true
}
],
"summary": "2 fixes needed: 1 critical, 1 high priority"
}
Development
# Start HTTP server (for programmatic clients)
bun server.mts
# or
bun run start
# Start stdio server (for Claude Desktop)
bun cli.mts
# or
bun run start:stdio
# Health check (HTTP mode only)
curl http://localhost:3000/health
# Set custom port (HTTP mode only)
MCP_PORT=8080 bun server.mts
# Type check
npx tsc --noEmit
Server Endpoints (HTTP Mode)
| Endpoint | Method | Description |
|---|---|---|
/health |
GET | Health check, returns { status: 'ok' } |
/mcp |
POST/GET/DELETE | MCP Streamable HTTP transport endpoint |
Architecture
server.mts # HTTP server (for programmatic clients)
cli.mts # Stdio server (for Claude Desktop)
browser/
├── driver.ts # Headless Chrome WebDriver
├── load-page.ts # Page loading, Core Web Vitals, screenshots
└── mod.ts
lib/
├── analysis/ # Text analysis & scoring algorithms
├── cache/ # Resource caching
├── config/ # Constants & configuration
├── extraction/ # Page parsing helpers
├── page/ # Page retrieval orchestration
├── suggestions/ # SEO improvement suggestions
├── tools/ # MCP tool implementations (24 tools)
└── types/ # TypeScript interfaces
Adding New Tools
- Create tool file in
lib/tools/your-tool.ts - Export registration function:
registerYourTool(server: McpServer) - Import and call in
lib/tools/mod.ts - Add to
registerAllTools()function
Technical Stack
- Runtime: Bun (primary), Node.js compatible
- HTTP Framework: Hono
- Transports: StdioServerTransport (Claude Desktop), WebStandardStreamableHTTPServerTransport (HTTP clients)
- Browser Automation: Selenium WebDriver with headless Chrome
- Key Dependencies:
@modelcontextprotocol/sdk,zod,hono,fast-xml-parser,node-html-parser
License
MIT
推荐服务器
Baidu Map
百度地图核心API现已全面兼容MCP协议,是国内首家兼容MCP协议的地图服务商。
Playwright MCP Server
一个模型上下文协议服务器,它使大型语言模型能够通过结构化的可访问性快照与网页进行交互,而无需视觉模型或屏幕截图。
Magic Component Platform (MCP)
一个由人工智能驱动的工具,可以从自然语言描述生成现代化的用户界面组件,并与流行的集成开发环境(IDE)集成,从而简化用户界面开发流程。
Audiense Insights MCP Server
通过模型上下文协议启用与 Audiense Insights 账户的交互,从而促进营销洞察和受众数据的提取和分析,包括人口统计信息、行为和影响者互动。
VeyraX
一个单一的 MCP 工具,连接你所有喜爱的工具:Gmail、日历以及其他 40 多个工具。
graphlit-mcp-server
模型上下文协议 (MCP) 服务器实现了 MCP 客户端与 Graphlit 服务之间的集成。 除了网络爬取之外,还可以将任何内容(从 Slack 到 Gmail 再到播客订阅源)导入到 Graphlit 项目中,然后从 MCP 客户端检索相关内容。
Kagi MCP Server
一个 MCP 服务器,集成了 Kagi 搜索功能和 Claude AI,使 Claude 能够在回答需要最新信息的问题时执行实时网络搜索。
e2b-mcp-server
使用 MCP 通过 e2b 运行代码。
Neon MCP Server
用于与 Neon 管理 API 和数据库交互的 MCP 服务器
Exa MCP Server
模型上下文协议(MCP)服务器允许像 Claude 这样的 AI 助手使用 Exa AI 搜索 API 进行网络搜索。这种设置允许 AI 模型以安全和受控的方式获取实时的网络信息。