Skip to content
Live diagnostic

Sitewell MCP Server

Give your AI assistant the ability to actually run a website health check — redirect chains, loops, bulk status, SSL expiry, broken links, headers and DNS. Free, no key.

Connect it

The server speaks the Model Context Protocol over streamable HTTP at a single endpoint. Point any MCP client at it and the tools appear.

https://sitewell.app/api/mcp

In Claude Code:

claude mcp add --transport http sitewell https://sitewell.app/api/mcp

In a client that takes a JSON config, the same thing:

{
  "mcpServers": {
    "sitewell": {
      "type": "http",
      "url": "https://sitewell.app/api/mcp"
    }
  }
}

There is no key to configure and no account to create. The endpoint is rate limited per IP, and every tool is a read — nothing it does can change the site being checked.

The seven tools

trace_redirects
Follow a URL's redirect chain hop by hop with per-hop response times, the total delay the redirects add, and severity findings for long chains, misused 302s, HTTPS-to-HTTP downgrades and www flips.
check_redirect_loop
Test a URL for an infinite loop, report the hop where the cycle closes, and name the pattern: HTTP/HTTPS ping-pong, www versus non-www, trailing slash, or a self-redirect.
check_http_status
Request up to 100 URLs and return each final status code, whether it redirected, where it landed, and how long it took.
check_ssl_expiry
Read the live certificate for a host: expiry date, days remaining, issuer — plus the domain registration expiry, which runs on a separate clock.
check_broken_links
Fetch a page, follow every outbound link, and report the ones that are broken with the status they returned.
check_http_headers
Return a URL's response headers, including the security and caching ones worth auditing: HSTS, CSP, Cache-Control, X-Content-Type-Options, Referrer-Policy.
dns_lookup
Resolve A, AAAA, MX, TXT, NS, CNAME and SOA records for a domain.

Why a tool call beats a guess

Ask an assistant why a URL loops and it can explain the mechanism perfectly and still not know whether yours does. Whether your chain drops to HTTP at hop two, how many milliseconds those hops cost, what date your certificate actually expires — none of that can be reasoned out. It requires a request to your server, right now.

That is the whole point of this endpoint. The assistant keeps the part it is good at, which is deciding what to check and what the answer means, and hands the measuring to something that measures. The same engines back the free web tools, so a result from the endpoint and a result from the site are the same result.

Frequently asked questions

What is the Sitewell MCP server?
It is an endpoint that lets an AI assistant run Sitewell's website health checks directly, instead of describing what a check would return. The Model Context Protocol is the open standard clients like Claude and ChatGPT use to call external tools, and Sitewell exposes seven of them over it: redirect tracing, loop detection, bulk status checking, SSL and domain expiry, broken links, response headers and DNS.
Do I need an account or an API key?
No. The endpoint is open and unauthenticated, the same as the tools on the site. It is rate limited per IP to keep it usable for everyone. Nothing you check is stored.
How do I add it to Claude?
Add it as a custom connector with the URL https://sitewell.app/api/mcp — Claude will discover the tools from the endpoint. In Claude Code, the equivalent is: claude mcp add --transport http sitewell https://sitewell.app/api/mcp
Which transport does it use?
Streamable HTTP, the current MCP transport. The server is stateless: each request is a self-contained JSON-RPC call, so there is no session to establish or keep alive and nothing to reconnect after an idle period.
Is it safe to point it at any URL?
The same guard that protects the web tools applies here: requests to private, loopback and link-local addresses are refused at every hop of a redirect chain, so the endpoint cannot be used to reach an internal network. Every check is a plain read — nothing the server does can modify the site being checked.
Why use this instead of asking the assistant to check a URL itself?
Because most of these answers cannot be reasoned out. Whether your chain drops to HTTP at hop two, how many milliseconds it costs, what your certificate's real expiry date is — those require an actual request to the actual server. This gives the assistant the measurement rather than an estimate of it.