WebDecoy / GEOMCP
Audit your site
for AI search.
Find technical SEO issues. Get evidence and practical fixes in your AI assistant.
Install GEOMCP
audit_url → evidence → fixThe output, before the pitch
See what needs fixing.
Explore real findings from our own site. Switch between the original result and the verified fix.
{
"url": "https://webdecoy.com/features/",
"id": "link-labels",
"category": "links",
"status": "warn",
"title": "Descriptive link labels",
"evidence": "16 links; 1 empty or generic labels.",
"recommendation": "Use link labels that communicate the destination or purpose.",
"weight": 1
}Recorded audit evidence, not a live URL scan. Fixes were rechecked on production. Inspect the changes.
Watch the 20-second walkthrough
An animated walkthrough of recorded MCP results. Read the transcript.
We used it on WebDecoy
193 pages.
11 things to improve.
The audit found skipped heading levels, unnamed checklist controls, a vague link label, and pages that needed useful internal links.
We fixed them, then used the same MCP to check the result. All 11 original warnings cleared on the live site.
Read the case studyOne site. Technical checklist results, not a ranking or citation study.
Start with one page
Your next audit
starts here.
Requires Node.js 22 or newer and an MCP-capable assistant. No account or API key.
This setup uses the built GitHub release. You do not need to clone or compile the project.
Release and checksumAdd this to .cursor/mcp.json in your project, then enable the server.
{
"mcpServers": {
"geomcp": {
"command": "npx",
"args": [
"--yes",
"--package=https://github.com/WebDecoy/geomcp/releases/download/v0.1.1/webdecoy-geomcp-0.1.1.tgz",
"geomcp"
]
}
}
}Run it from the terminal
npx --yes --package=https://github.com/WebDecoy/geomcp/releases/download/v0.1.1/webdecoy-geomcp-0.1.1.tgz geomcpThe command starts the stdio server and waits for an MCP client. Add --help to see command help.
Then ask: “Audit https://my-site.com. Show the three most useful fixes and the evidence for each.”
Five tools. A complete audit loop.
Inspect. Improve. Recheck.
| Tool | What you get |
|---|---|
audit_url | A public page audit with evidence, coverage, and prioritized fixes. |
audit_html | An offline check of supplied HTML, including pre-deployment output. |
inspect_robots | Declared search, retrieval, and training policies for a page path. |
inspect_sitemap | One sitemap's URLs, duplicates, date syntax, and discovery hints. |
compare_audits | What improved, regressed, or could not be evaluated. |
26 checks cover discovery, HTML structure, structured data, media, links, evidence signals, and delivery. Read every check.
Good questions.
Straight answers.
Does this improve my AI search ranking?
It helps you identify technical issues that affect how your page can be discovered and understood. Its score does not predict rankings, AI citations, factual accuracy, or signups.
Does my page get sent to a third-party model?
GEOMCP runs locally and does not call a model API or send telemetry. Results go to your MCP client. Your assistant provider's own data policies still apply.
Will it edit my site or block crawlers?
No. Its tools are read-only. Your assistant can propose edits using the findings; GEOMCP itself does not change files, deploy code, or enforce crawler rules.
Can it see JavaScript-rendered content?
It inspects server-delivered HTML, not a rendered browser. Use audit_html with HTML you supply when that is more appropriate. It does not measure Core Web Vitals.
Is this part of a paid WebDecoy plan?
No. GEOMCP is an independent MIT-licensed project by WebDecoy. You can use it without signing up for WebDecoy.
Also building something people can visit?
See who actually
visits your site.
WebDecoy helps you inspect bot traffic and suspicious activity. Watch a real controlled browser trial, then connect your own site.
See bot detection in action