I Built an MCP Server with 36 Developer Tools. Now Claude Does My Site Audits.

#ai #mcp #security #showdev

I got tired of juggling browser tabs to check SSL certs, DNS records, security headers, and broken links. So I built an MCP server that lets Claude do it all.

One config block. 36 tools. Here's what that actually looks like.

"Run a security audit on mysite.com"

That's it. That's the prompt. Claude calls three APIs behind the scenes: SSL certificate check, security headers analysis, and email authentication (SPF, DKIM, DMARC). Ten seconds later I get a report with actual findings.

I didn't build any orchestration logic. I gave Claude atomic tools and it figured out how to chain them together.

Some other things I've been using this for:

Bulk SSL monitoring:
"Check SSL certificates for stripe.com, github.com, and vercel.com. Flag anything expiring within 30 days."

Claude loops through the list, calls check_ssl for each, compares expiry dates, and gives me a summary table.

Screenshot comparison in Claude Code:
"Take a screenshot of our staging site and production site. Tell me if anything looks different."

Claude captures both screenshots, then uses its vision to spot visual differences. No Playwright setup, no test scripts. Just a sentence.

Email list cleaning:
"Read contacts.csv, validate every email in the 'email' column, create a new file with only the valid ones."

Claude reads the file, calls validate_email per row, writes the clean output. Took maybe 15 seconds for 200 rows.

DNS migration check:
"We moved mysite.com to 203.0.113.50. Has the DNS propagated?"

Claude calls dns_lookup, checks the A record, tells me if it matches.

Setup takes 30 seconds

Add this to your MCP client config:

{
  "mcpServers": {
    "apixies": {
      "command": "npx",
      "args": ["-y", "@apixies/mcp-server"],
      "env": {
        "APIXIES_API_KEY": "your-key-here"
      }
    }
  }
}

Config file locations:

Claude Desktop (macOS): ~/Library/Application Support/Claude/claude_desktop_config.json
Claude Desktop (Windows): %APPDATA%\Claude\claude_desktop_config.json
Cursor: .cursor/mcp.json
VS Code: .vscode/mcp.json

Restart your client. 36 tools show up. Done.

Free API key at apixies.io/register. Works without one too (sandbox mode, 20 requests/day).

The full tool list

Inspectors: SSL checker, security headers, DNS lookup, email inspector, WHOIS, link checker, redirect tracer, meta tag extractor, robots.txt parser, favicon fetcher, URL validator, email auth validator, user agent parser, IP geolocation, website performance, JWT decoder.

Converters: HTML to PDF, Markdown to HTML, JSON formatter, JSON to CSV, Base64 encode/decode, URL encode/decode, color converter, timestamp converter, phone formatter, JSON schema validator.

Generators: QR code, screenshot, password, UUID, hash, lorem ipsum, URL shortener.

Each one has a description that tells the AI when to use it. You don't pick tools manually. You describe what you want and Claude figures out which ones to call.

Prompts worth trying

Start simple:

"Check the SSL certificate for github.com"
"What DNS records does google.com have?"
"Generate a 24-character password with symbols"

Then try the multi-tool stuff:

"Check SSL, headers, and performance for these 5 client sites. Give me a comparison table."
"Take a screenshot of example.com and check if their OpenGraph meta tags are set up correctly."
"Decode this JWT and tell me when it expires: eyJhbG..."
"Scan all markdown files in ./docs/ for URLs and check if any are broken."

That last one is my favorite. Claude finds every URL in your docs, calls the link checker for each, and reports which ones are dead. Would've taken me an hour to write as a script.

How it works (briefly)

MCP is an open protocol from Anthropic. You define tools with a name, description, and parameter schema. The AI reads the descriptions and calls them when relevant.

The server is a small TypeScript package. Each of the 36 API endpoints is registered as a tool with a Zod schema for input validation. When Claude calls a tool, the server makes an HTTP request to the Apixies API and returns the JSON response.