Google Search Console MCP,
for your AI agents.
The hosted, read-only Search Console MCP server that brings your real search performance data into Claude, ChatGPT, Cursor and more — connect in about a minute, no install, no API keys.
Set it up in your AI
Pick your client — the steps are below. Same endpoint everywhere.
- Open Settings → Connectors in Claude Desktop or claude.ai.
- Click Add custom connector and paste the URL below.
- Click Add, then sign in with Google when prompted.
https://mcpsearchconsole.com/mcp- Add the server from your terminal:
claude mcp add --transport http search-console https://mcpsearchconsole.com/mcp
- Run /mcp inside Claude Code and complete Google sign-in.
- Go to Settings → Connectors and enable Developer mode.
- Click Create / Add custom MCP server and paste the URL below.
- Authenticate with Google to finish.
https://mcpsearchconsole.com/mcp- Add this to ~/.codex/config.toml:
[mcp_servers.search-console] url = "https://mcpsearchconsole.com/mcp"
On Google sign-in, Codex opens your browser to authorize. (If your Codex version only supports stdio servers, bridge it with npx mcp-remote https://mcpsearchconsole.com/mcp.)
- In Mistral Vibe → Settings → Connectors, choose Add MCP connector.
- Paste the URL below and save.
- Sign in with Google when prompted.
https://mcpsearchconsole.com/mcp- Add to ~/.cursor/mcp.json (or Settings → MCP → Add server):
{
"mcpServers": {
"search-console": { "url": "https://mcpsearchconsole.com/mcp" }
}
}
Reload, then authorize with Google from the MCP panel.
Any client that supports remote MCP servers over Streamable HTTP. Point it at the URL — it handles Google sign-in via OAuth.
{
"mcpServers": {
"search-console": { "type": "streamable-http", "url": "https://mcpsearchconsole.com/mcp" }
}
}Connect Search Console to AI in three steps
A remote MCP connector for Google Search Console — point your client at one URL and go.
Pick your AI client
Use the setup tabs above for Claude, Claude Code, ChatGPT, Codex, Mistral Vibe, Cursor and more.
Add the MCP endpoint
Drop the hosted MCP URL into your client and authorize with Google when prompted.
Ask about your traffic
"Which queries are rising?" "What pages lost clicks?" Your AI answers from your real GSC data.
What your AI can do with it
A focused Search Console connector — the data that actually answers SEO questions.
List your properties
Every Search Console site your Google account can access, read-only.
Search analytics
Clicks, impressions, CTR and position — filtered by query, page, country or device, with period-over-period comparison.
Inspect & diagnose
Check any URL's index status, coverage, canonical and last crawl — plus the sitemaps you've submitted and their errors.
Speed & site health
Core Web Vitals and PageSpeed, redirect chains, robots.txt crawlability, DNS and TLS certificates — the technical health behind your rankings.
Find opportunities
Striking-distance keywords, rising and falling pages, low-CTR wins — analyst work, not raw dumps.
Works everywhere
Remote MCP over Streamable HTTP — Claude, Cursor, Windsurf, ChatGPT. No install.
The tools it exposes
Exactly what the MCP server offers your AI — read-only, no write access.
list_sites
Every Search Console property your Google account can access.
search_analytics
Clicks, impressions, CTR and position — filter by query/page/country/device and compare periods.
inspect_url
Index status, coverage, last crawl, canonical and mobile checks for any URL.
list_sitemaps
Submitted sitemaps with status, errors and indexed counts.
check_dns
DNS records for any domain — A, MX, TXT and more, for ownership and site checks.
page_speed
Core Web Vitals & PageSpeed — real-user LCP/INP/CLS plus Lighthouse score and top fixes.
check_redirects
Trace a URL's redirect chain — every hop, the final destination, loops and chains.
check_robots
Is this URL crawlable? robots.txt rules, the matching directive and declared sitemaps.
check_tls
TLS certificate health — issuer, expiry, hostname match and chain validity.
Why a hosted Search Console MCP
- Hosted, not self-host: skip the Python/Node + Google Cloud setup that open-source GSC MCP servers require.
- Read-only by design: the least-privilege scope means this SEO MCP can read your data but never touch your site.
- One AI connection to Search Console: the same endpoint works across every MCP-capable client.
Hosted MCP vs local stdio MCP
Open-source Search Console MCP servers run locally over stdio. This one is hosted and remote (Streamable HTTP).
| This hosted MCP | Local stdio MCP | |
| Transport | Remote · Streamable HTTP (URL) | Local · stdio (subprocess) |
| Setup | Sign in with Google, paste a URL | Install Python/Node, clone a repo |
| Google Cloud project | Not needed | You create one + OAuth credentials |
| Credentials | OAuth, managed for you | You manage a client secret / JSON key |
| Web & mobile clients | Yes (claude.ai, ChatGPT, Mistral Vibe) | No — desktop/CLI only |
| Updates | Automatic | You pull & redeploy |
| Access | Read-only by design | Depends on the repo's scopes |
Frequently asked questions
Google Search Console MCP, hosted connectors and remote MCP servers.
What is a Google Search Console MCP?
It is an MCP (Model Context Protocol) server that connects Google Search Console to AI clients. Once added, your AI can read your Search Console performance data — clicks, impressions, CTR and average position by query, page, country and device — and answer questions about your SEO in plain language.
Is it read-only and safe?
Yes. We request only the read-only Search Console scope, so the connector can read your performance and index data — analytics, URL inspection and your submitted sitemaps — but can never modify your site, submit sitemaps or change settings. Your Google sign-in only ever exposes the properties you already own. The technical-health tools (DNS, PageSpeed, redirects, robots.txt, TLS) only read public web data and need no access to your account at all.
Which AI clients are supported?
Any client that supports remote MCP servers over Streamable HTTP — including Claude (Desktop, Code and claude.ai), Cursor, Windsurf and ChatGPT. You add one URL and sign in with Google; no install and no API keys.
What is a hosted / remote MCP server (Streamable HTTP)?
A remote MCP server runs in the cloud and is reached by URL over the Streamable HTTP transport, instead of being installed and run on your machine. This hosted Search Console connector means there is nothing to deploy — you just point your AI client at the endpoint.
Hosted MCP vs a local (stdio) MCP server — what is the difference?
Open-source Search Console MCP servers run on your machine and talk to the AI over stdio, which means installing Python or Node, creating a Google Cloud project and managing OAuth credentials yourself, and they only work in desktop or CLI clients. This hosted connector is a remote MCP server over Streamable HTTP: you sign in with Google and paste a URL, nothing to install, and it works in web and mobile clients too.
How do I connect Search Console to AI?
Sign in with Google, add the MCP endpoint to your AI client, and ask about your search traffic. The whole setup takes about a minute.
Put your search data in your AI
Free to connect · read-only · about a minute to set up.