MCP server

Agents run the same public checkers the website exposes. No account. This page is the documentation; the endpoint below is the protocol.

Endpoint

https://api.stillindexed.com/mcp

Streamable HTTP, stateless JSON-RPC. GET returns server info. POST is initialize, tools/list and tools/call.

Install

Claude Code

claude mcp add --transport http stillindexed https://api.stillindexed.com/mcp

Cursor

Put this in .cursor/mcp.json or Cursor Settings → MCP.

{
  "mcpServers": {
    "stillindexed": {
      "url": "https://api.stillindexed.com/mcp"
    }
  }
}

Install in Cursor (deep link, same config).

Codex CLI

codex mcp add stillindexed --url https://api.stillindexed.com/mcp

ChatGPT and other agents

Paste https://api.stillindexed.com/mcp as a remote MCP server URL. Remote HTTP, not stdio; no key.

Tools

Tool	Does	Budget
stillindexed_check_url	Indexability: status, redirects, meta robots, X-Robots-Tag, canonical, title, H1	1 fetch
stillindexed_bulk_check	Same check, up to 20 URLs	1 fetch each
stillindexed_check_robots	Effective allow/disallow per path, wildcard vs Googlebot	1 fetch
stillindexed_check_sitemap	Status, XML validity, URL count	1 fetch
stillindexed_check_tls	Certificate days remaining, hostname match	1 fetch
stillindexed_check_domain	RDAP expiry. Silent if the TLD has no RDAP	1 fetch
stillindexed_crawler_diff	Googlebot, Bingbot, GPTBot and a browser, compared	4 fetches
stillindexed_generate_seo_skill	SKILL.md from the page, robots.txt, sitemap (index expanded), host variants, crawler diff and TLS	up to 12

Shared cap: 60 outbound fetches per IP per hour, the same bucket asthe website tools. Repeat checks of one URL are cached for 60 seconds and do not spend the budget.

Rate limit and safety

The SSRF guard refuses loopback, private ranges and link-local addresses — the same guard the website tools use, enforced at dial time, so a redirect cannot bypass it. No JavaScript rendering, no rank lookups.

Example

Ask the agent: “Is https://example.com/pricing indexable? Check meta robots and X-Robots-Tag.”

It should call stillindexed_check_url with that URL and readindexable, directives.meta_robots anddirectives.x_robots_tag.

“Write an SEO skill for https://example.com” should callstillindexed_generate_seo_skill and save the returnedskill_markdown as SKILL.md. Same generator as thepublic tool.

Questions

Does the MCP server need an API key?

No. It is the same ungated checkers as the website tools. Rate limit is 60 outbound fetches per IP per hour, shared with stillindexed.com/tools.

What can an agent check?

One URL for indexability (meta robots, X-Robots-Tag, canonical, redirects, title, H1), bulk of up to 20 URLs, robots.txt per path, sitemap validity, TLS expiry, domain expiry via RDAP, and crawler-versus-browser diffs (Googlebot, Bingbot, GPTBot, browser).

Will it fetch localhost or cloud metadata?

No. The same SSRF guard as the website tools refuses loopback, private ranges and link-local addresses.

How many fetches does the skill generator spend?

Up to 12: the page, robots.txt, the sitemap (plus the first child when it is a sitemap index), three host/scheme variants, a four-persona crawler diff, and the TLS handshake. One tool call, one budget.

Directories

Submit https://stillindexed.com/mcp as the listing URL andhttps://api.stillindexed.com/mcp as the server URL.

Generate an SEO skill for your host ·Crawler transparency