Skip to main content
Glama
@operstack/mcp

@operstack/mcp

Official
by oper-stack

@operstack/mcp

Let Claude, Cursor or any MCP client measure a site the way a search engine and an AI actually read it.

Ask your assistant "why does ChatGPT recommend my competitor and not me?" and it can now go and look: read the site, score six areas, follow the map the site offers to machines, and tell you what is actually missing. Free public signals only. No account, no API key, nothing stored.

Install

In Claude Desktop, add this to claude_desktop_config.json:

{
  "mcpServers": {
    "operstack": {
      "command": "npx",
      "args": ["-y", "@operstack/mcp"]
    }
  }
}

In Cursor, add the same block to .cursor/mcp.json in your project, or to ~/.cursor/mcp.json for every project.

Node 20 or newer. Nothing else to install.

Related MCP server: librecrawl-technical-seo-audit-mcp

What it can do

audit_site reads a public site and scores six areas out of ten: technical SEO, content and structure, AEO (whether an answer engine can quote it), GEO (whether AI systems can identify and use it), off-page trust, and conversion. Returns every failing and borderline check with what was actually found on the pages, so the assistant can explain rather than guess.

check_llms_txt reads the site's /llms.txt, checks it against the format, and follows its links to see whether each one still leads to a page, a redirect or nothing. This file rots silently, because no human ever opens it.

compare_sites runs the same measurement on two to four sites and returns them side by side. Useful when the question is not "am I bad" but "am I worse than them, and where".

run_gates runs the sixteen OperStack content gates on a folder of Markdown or MDX on your own machine: cut titles, copied paragraphs, hollow sections, dead links, redirect chains, a stale agent index, figures with no source, and the agent surface. Files are read locally and sent nowhere.

What it will not do

It will not invent a number. Where something was not measured it says so and why. It will not score a site on data that costs money: every figure comes from what any stranger can read on the site for nothing, which means you can reproduce it yourself. It will not report on a site that did not answer: a typo in the address gets a refusal, not a page of findings about nothing.

It keeps nothing. The address you check is used for the request and forgotten. There is no telemetry, no cache and no account.

Why the numbers match the paid report

Every measurement here runs through @operstack/audit, the same package behind the paid OperStack reports. A free tool that disagrees with the paid one is the fastest way to lose the reader, so they are the same code.

MIT.

Available Tools

4 tools
audit_siteAudit a site the way a search engine and an AI read itA

Read a public site and score six areas out of ten: technical SEO, content and structure, AEO (whether an answer engine can quote it), GEO (whether AI systems can identify and use it), off-page trust, and conversion. Returns every failing and borderline check with what was actually found on the pages. Free public signals only: no account, no paid tool, nothing stored. Use it when someone asks why a site is not being cited, why AI assistants recommend competitors, or what to fix first.

Inputs: url (required), lang, pages.

check_llms_txtCheck the llms.txt map a site offers to AIA

Read a site's /llms.txt, check it against the format (one H1 naming the site, a summary, sections of links) and follow its links to see whether each still leads to a page, a redirect or nothing. This file rots silently because no human ever opens it. Use it when someone asks whether their llms.txt is correct, why an assistant quotes the wrong pages, or before publishing a new one.

Inputs: url (required), maxLinks.

compare_sitesCompare a site with its rivals on the same measurementsA

Run the same six-area measurement on two to four public sites and return them side by side. Use it when someone asks why a competitor is being quoted instead of them, or wants to know where the gap actually is rather than guessing. Free public signals only.

Inputs: urls (required), lang, pages.

run_gatesRun the sixteen OperStack content gates on a local projectA

Run the sixteen quality gates on a folder of Markdown or MDX content on this machine: cut titles, copied paragraphs, hollow sections, dead links, redirect chains, a stale agent index, figures with no source, and the agent surface. Returns the report. Use it before publishing, or when asked what is wrong with a content corpus. It needs a local path, not a URL, and it reads files without sending them anywhere.

Inputs: path (required), only.

Tool Schema Changelog

Recent tool additions, removals, and schema changes observed during successful MCP inspections.

Loading tool schema changelog…

    TDQS

    A4/5.0

    Scored across 4 tools

    Disambiguation4/5

    Each tool has a fairly distinct purpose: single-site audit, multi-site comparison, llms.txt validation, and local content gates. However, compare_sites is essentially a multi-site extension of audit_site, and check_llms_txt overlaps with the AEO/GEO portion of audit_site, so an agent could occasionally waver between them despite the clarifying descriptions.

    Naming Consistency5/5

    All four tools follow a clean verb_noun snake_case pattern: compare_sites, audit_site, check_llms_txt, run_gates. The convention is predictable throughout with no style mixing.

    Tool Count4/5

    Four tools is a reasonable, well-scoped count for a site-measurement server, and each earns its place. It leans slightly thin, since there is no tool for remediation or reporting beyond the core diagnostics.

    Completeness4/5

    The surface covers the core diagnostic lifecycle: single-site audit, competitor comparison, a specific llms.txt check, and local corpus gates. Minor gaps remain—no historical tracking, export/reporting, or standalone robots/schema checks—but agents can work around these for the stated purpose.

    Related MCP Connectors

    Related MCP Servers

    • A
      license
      A
      quality
      B
      maintenance
      AutomateLab AI-SEO audits, scores, and rewrites web pages for AI citation eligibility, AEO and GEO. No API keys or registration. Works with Claude, Cursor, Codex, and other MCP clients. Product and documentation: https://automatelab.tech/products/mcp/ai-seo/
      20
      88 npm
      3
      MIT
    • A
      license
      Not graded
      quality
      B
      maintenance
      Open-source technical SEO crawler MCP server built on LibreCrawl. Runs full audits inside Claude, Cursor, or Codex — 50+ checks (hreflang, schema.org, security headers, WAF detection on 200-OK pages), chunked-progressive engine for large sites, ephemeral by design (server forgets every audit after download).
      47
      MIT
    • A
      license
      A
      quality
      C
      maintenance
      Enables AI agents to check whether a public website is crawlable, understandable, and ready for AI search workflows through local-only audits of robots.txt, sitemaps, metadata, and llms.txt.
      3
      73 npm
      2
      MIT
    • A
      license
      Not graded
      quality
      D
      maintenance
      Open-source MCP server that audits websites for AI search readiness, providing deterministic scoring (0-100) and prioritized fix lists for metrics like JSON-LD, llms.txt, heading hierarchy, and AI crawler access.
      1
      MIT