@mcpx-digital/site-health
# @mcpx-digital/site-health
**MCP server for polite, permission-based site health checks.**
Check HTTP status codes, find broken same-origin links, review SEO basics, and inspect common security headers — from Cursor, Claude Desktop, or any MCP client.
> **Permission required.** Only use this server on sites **you own** or have **explicit permission** to test. All tools are **defensive / informational only**. This package does **not** include exploit payloads, attack procedures, SQLi/XSS scanners, or other offensive tooling.
## Install (MCPX / npx)
```bash
npx -y @mcpx-digital/site-health
```
> **Publish note:** The package is prepared for npm as `@mcpx-digital/site-health`. If it is not on the registry yet, publish from this repo (`npm publish --access public`) after logging into an account with access to the `@mcpx-digital` scope. Until then you can run from a local clone or GitHub.
Fallback package name (if the `@mcpx-digital` scope is unavailable): `@theoryofshadows/site-health-mcp`.
## Cursor `mcp.json` example
Add to your Cursor MCP config (e.g. `~/.cursor/mcp.json` or project `.cursor/mcp.json`):
```json
{
"mcpServers": {
"site-health": {
"command": "npx",
"args": ["-y", "@mcpx-digital/site-health"]
}
}
}
```
Local clone (no npm publish required):
```json
{
"mcpServers": {
"site-health": {
"command": "node",
"args": ["/absolute/path/to/site-health-mcp/index.js"]
}
}
}
```
## Tools
| Tool | What it does |
|------|----------------|
| `http_status` | Check status codes for a list of URLs (max 25), with polite delay |
| `broken_links` | Same-origin crawl from a start URL, **max depth 2**, polite delay; report 4xx/5xx |
| `seo_basics` | Title, meta description, canonical present? + simple length / H1 hints |
| `security_headers` | Checklist of CSP, HSTS, X-Frame-Options, etc. — **presence/absence only** |
### Hard limits (built in)
- HTTP/HTTPS only
- Max depth **2** for crawls; max **40** pages
- Polite default delay (~400ms) between requests
- Custom User-Agent identifying this tool
- No paid APIs — plain Node.js `fetch` + `@modelcontextprotocol/sdk`
## Example prompts
- “Check HTTP status for https://example.com and https://example.com/missing”
- “Crawl https://mysite.com for broken links (depth 2)”
- “Run SEO basics on https://mysite.com/blog/post”
- “Which security headers are present on https://mysite.com?”
## Development
```bash
git clone https://github.com/TheoryofShadows/site-health-mcp.git
cd site-health-mcp
npm install
npm test
node index.js # stdio MCP server
```
## Sell / list on MCPX
Suggested listing price: **$5–$9**.
1. Ensure the package is published to npm (`@mcpx-digital/site-health`).
2. List on [MCPX](https://www.mcpx.digital) with install command:
`npx -y @mcpx-digital/site-health`
3. Emphasize: permission-based, defensive only, zero paid APIs, MIT.
## License
MIT © TheoryofShadows
TDQS
Scored across 4 tools
http_status and broken_links both report HTTP status codes, but one targets a provided list while the other crawls same-origin; descriptions clarify the boundary. seo_basics and security_headers are clearly distinct from each other and the other tools.
All tools use consistent snake_case noun phrases (http_status, broken_links, seo_basics, security_headers) with no mixing of conventions. The pattern is predictable and readable.
Four focused tools exactly match the site-health domain, each covering a distinct diagnostic area. No tool feels redundant or out of place for the stated purpose.
The set covers status codes, broken links, SEO basics, and security headers, which are core site health checks. However, it lacks common diagnostics like redirect chains, SSL/TLS certificate validity, performance, or accessibility—minor gaps for a comprehensive suite.