Crawlinx MCP

Official5 toolsLive

by im-ff

Free technical-SEO audit MCP: crawl a site, run checks, return an LLM-ready shareable report.

Topics

ai-agentsclaudemcpmodel-context-protocolseoseo-audittechnical-seoweb-crawler

Captured live from the server via tools/list.

scan_url

Submit a URL to Crawlinx for an SEO audit. Returns a scan id; poll get_report until the crawl finishes.

Parameters2

mode	string	optional	Scan mode: "free" (default) or "instant".
url	string	required	The absolute http(s) URL to audit.

Raw schema

{
  "type": "object",
  "properties": {
    "mode": {
      "description": "Scan mode: \"free\" (default) or \"instant\".",
      "type": "string"
    },
    "url": {
      "description": "The absolute http(s) URL to audit.",
      "type": "string"
    }
  },
  "required": [
    "url"
  ]
}

get_report

PREFERRED for reasoning about what to fix. Fetch a finished report as structured JSON (health score, category subscores, all issues with URLs, stats). Use this when you need to decide which problems matter.

Parameters1

scan_id

string

required

The scan id returned by scan_url.

Raw schema

{
  "type": "object",
  "properties": {
    "scan_id": {
      "description": "The scan id returned by scan_url.",
      "type": "string"
    }
  },
  "required": [
    "scan_id"
  ]
}

get_report_markdown

Fetch a finished report as compact, agent-ready Markdown: health, a prioritized fix plan with remediation, and issues collapsed to one row per code (code | severity | count | example URLs) — small enough to keep in context on a real site. Set full=true for the complete human report (large). Use get_issue_detail to expand any single code.

Parameters3

domain	string	optional	Optional domain label for the report heading.
full	boolean	optional	When true, return the full human report (all occurrences, checks coverage). Default false = compact agent view.
scan_id	string	required	The scan id returned by scan_url.

Raw schema

{
  "type": "object",
  "properties": {
    "domain": {
      "description": "Optional domain label for the report heading.",
      "type": "string"
    },
    "full": {
      "description": "When true, return the full human report (all occurrences, checks coverage). Default false = compact agent view.",
      "type": "boolean"
    },
    "scan_id": {
      "description": "The scan id returned by scan_url.",
      "type": "string"
    }
  },
  "required": [
    "scan_id"
  ]
}

get_issue_detail

Drill into one issue code: every affected URL plus step-by-step remediation (what to change, why, and when it's a false alarm). Call this after get_report_markdown to expand a code from the collapsed table before applying fixes.

Parameters2

code	string	required	The issue code to expand, e.g. "links.broken_internal" or "title.too_long".
scan_id	string	required	The scan id returned by scan_url.

Raw schema

{
  "type": "object",
  "properties": {
    "code": {
      "description": "The issue code to expand, e.g. \"links.broken_internal\" or \"title.too_long\".",
      "type": "string"
    },
    "scan_id": {
      "description": "The scan id returned by scan_url.",
      "type": "string"
    }
  },
  "required": [
    "scan_id",
    "code"
  ]
}

diff_scan

Diff a scan against the previous scan of the same domain: health movement plus the issue codes that were fixed or newly appeared/regressed. Use this when auditing in a loop to verify fixes landed and nothing regressed.

Parameters1

scan_id

string

required

The current scan id; the previous scan of the same domain is found automatically.

Raw schema

{
  "type": "object",
  "properties": {
    "scan_id": {
      "description": "The current scan id; the previous scan of the same domain is found automatically.",
      "type": "string"
    }
  },
  "required": [
    "scan_id"
  ]
}

README.md3 KBView on GitHub

Crawlinx MCP

Free technical-SEO audit as an MCP server — built for AI coding agents.

Crawlinx is a hosted Model Context Protocol server that lets your AI agent audit a website's technical SEO and act on the results. It crawls a site (up to 100 pages), runs 77 technical-SEO checks, and returns an LLM-ready Markdown report with a prioritized fix plan — small enough to keep in context on a real site.

☁️ Remote server over Streamable HTTP — no local install, no API key, no signup
🔎 77 checks — titles, meta, canonicals, orphan pages, noindex, internal PageRank, JS render-diff, images, performance, sitemap
🤖 Agent-native — compact reports, per-issue drill-down, and scan diffs so an agent can scan → fix → verify in a loop
🔗 Public shareable reports — every scan has a URL

Homepage: https://crawlinx.com · Docs: https://crawlinx.com/api

Install

The server is remote — you only need the endpoint https://crawlinx.com/mcp. No key.

Claude Code

bash

claude mcp add --transport http crawlinx https://crawlinx.com/mcp

Cursor / Windsurf / Cline / Claude Desktop

Add to your MCP config (mcp.json / settings):

json

{
  "mcpServers": {
    "crawlinx": {
      "type": "http",
      "url": "https://crawlinx.com/mcp"
    }
  }
}

A machine-readable manifest is served at /.well-known/mcp.json.

Tools

Tool	What it does
`scan_url`	Start a free SEO audit of a URL; returns a scan id.
`get_report`	Fetch a finished report as structured JSON.
`get_report_markdown`	Fetch a finished report as compact, LLM-ready Markdown (health score + prioritized fix plan + issues collapsed by code). Set `full=true` for the complete report.
`get_issue_detail`	Expand one issue code: every affected URL plus step-by-step remediation.
`diff_scan`	Diff a scan against the previous scan of the same domain to confirm fixes landed and nothing regressed.

Example prompts

"Audit https://example.com and give me the top 5 SEO fixes."
"Scan my site, then fix the duplicate <title> issues in the codebase and re-scan to confirm."
"Which pages are orphaned (no internal links) and how do I link them?"

Typical workflow

scan_url → get a scan_id
poll get_report_markdown until the crawl finishes
get_issue_detail(scan_id, code) to expand a specific issue before changing code
apply fixes
re-scan, then diff_scan(scan_id) to verify

The crawler

Audits run as CrawlinxBot, which honors robots.txt, limits request rates, and does not bypass access controls. See crawlinx.com/bot. Site owners can opt out via robots.txt or the removal form.

About

This repository documents the hosted Crawlinx MCP server. The audit engine is operated as a service at crawlinx.com; this repo is the public reference and install guide for the MCP integration.

Install

Tools (5)

scan_urlget_reportget_report_markdownget_issue_detaildiff_scan

Remote endpoint

Streamable HTTP

Hosted server - connect over the network, no local install.

url

https://crawlinx.com/mcp

Add to VS Code Add to Cursor

claude_desktop_config.json

json

{
  "mcpServers": {
    "crawlinx-mcp": {
      "command": "npx",
      "args": [
        "-y",
        "mcp-remote",
        "https://crawlinx.com/mcp"
      ]
    }
  }
}

Desktop config is stdio-only; this bridges via mcp-remote. Native remote: Settings > Connectors.