Google Search Console MCP Server

Google Search Console MCP Server

A Model Context Protocol (MCP) server that gives AI agents direct access to your Google Search Console data.

• Search analytics — query clicks, impressions, CTR, and average position by page, query, country, device, or date range
• Compare periods — week-over-week, month-over-month trends
• Find opportunities — high-impression/low-click queries, ranking keywords you didn't know about
• Track specific pages — see which URLs are gaining or losing traction
• Index coverage — check which pages are indexed, excluded, or erroring
• Sitemap status — verify sitemaps are being read and how many URLs are indexed

Read-only access — this server cannot submit URLs, modify settings, or make any changes to your Search Console properties.

[!IMPORTANT] Two authentication methods are supported — signing in as yourself with Application Default Credentials, or a service account key. See Authentication for the trade-offs.

Tools

List all sites (properties) you have access to in Google Search Console.

No parameters required.

Query search analytics data — clicks, impressions, CTR, and position.

Check indexing status, crawl info, and mobile usability for a URL.

List all submitted sitemaps and their status for a site.

Installation

Quick setup via agent

Already signed in with gcloud? Paste this prompt into Claude Code or Claude Desktop and it will configure everything for you:

Clone and build the Google Search Console MCP server from https://github.com/sarahpark/google-search-console-mcp, then add it to my global MCP config. After setup, call list_sites to verify it works.

If you are using a service account key instead, add "My service account key is at /path/to/service-account-key.json" to the prompt and replace the path with your actual file location.

Build from source

git clone https://github.com/sarahpark/google-search-console-mcp.git
cd google-search-console-mcp
npm install
npm run build

Configuration

Google Cloud project

1. Go to Google Cloud Console and create a new project (or select an existing one)
2. Open the Search Console API page and click Enable

Authentication

Pick one of the two options below. The server uses the standard Google auth chain, so it picks up whichever you set up: the service account key at GOOGLE_APPLICATION_CREDENTIALS if that variable is set, otherwise your gcloud Application Default Credentials.

Option A: sign in as yourself (ADC)

1. Sign in, requesting the Search Console scope:

gcloud auth application-default login \
  --scopes=https://www.googleapis.com/auth/webmasters.readonly,https://www.googleapis.com/auth/cloud-platform

1. Set the quota project:

gcloud auth application-default set-quota-project YOUR_PROJECT_ID

Your credentials must end up with a quota project — with user credentials and no quota project attached, the Search Console API returns 403 PERMISSION_DENIED:

Your application is authenticating by using local Application Default Credentials. The searchconsole.googleapis.com API requires a quota project, which is not set by default.

Step 1 tries to attach your current gcloud config project automatically, but it silently skips that when the account you signed in as lacks the serviceusage.services.use permission on it (project Editor and Owner both include that permission), leaving you with credentials that get the 403 above:

WARNING: Cannot add the project "YOUR_PROJECT_ID" to ADC as the quota project because the account in ADC does not have the "serviceusage.services.use" permission on this project.

Running step 2 explicitly is the reliable path, and it is also how you bill a project other than your configured one. The cloud-platform scope in step 1 is what allows the quota project to be attached.

You do not need to create your own OAuth client ID or pass --client-id-file: gcloud's built-in client grants webmasters.readonly (confirmed on Google Cloud SDK 557.0.0). The gcloud reference does say to create an OAuth client ID for scopes outside Google Cloud, so if a future version stops granting this one, that is the fallback:

gcloud auth application-default login \
  --client-id-file=client_id.json \
  --scopes=https://www.googleapis.com/auth/webmasters.readonly

Because the calls run as you, nothing needs to be added in Search Console — you already have access to every property you own or have been granted.

Option B: service account key

1. In the Google Cloud Console sidebar, go to APIs & Services → Credentials
2. Click Create Credentials → Service account
3. Give it a name (e.g. "search-console-mcp"), then click Create and Continue
4. You can skip the optional role/access steps — click Done
5. On the Credentials page, click on the service account you just created
6. Go to the Keys tab → Add Key → Create new key → select JSON → click Create
7. A .json key file will download — save it somewhere safe (e.g. ~/.config/gcloud/service-account-key.json)

[!CAUTION] Treat this key file like a secret key. Do not save it inside your project repo or commit it to git. Store it outside your project directory and reference it by absolute path in your MCP config.

Then add the key path to the server's environment in whichever config you use below:

      "env": {
        "GOOGLE_APPLICATION_CREDENTIALS": "/absolute/path/to/service-account-key.json"
      }

The service account also has to be granted access to each property you want to read:

1. Copy the service account's email address (it looks like name@project-id.iam.gserviceaccount.com — you can find it on the service account details page)
2. Go to Google Search Console
3. Select your property, then go to Settings → Users and permissions
4. Click Add user, paste the service account email, set the permission to Restricted, and click Add

[!WARNING] Step 4 currently fails with "Failed to add user: email not found" for newly created service accounts — this is what motivated adding Option A. If you hit it, use Option A, or use an existing service account that was already added to the property.

Claude Code

claude mcp add gsc --scope user -- node /absolute/path/to/google-search-console-mcp/build/index.js

--scope user makes the server available in all your projects. Use --scope project instead to write it to .mcp.json in the current directory and share it with your team. Omitting --scope defaults to local, which only enables it for you in the current project.

Or add it manually to the "mcpServers" object in ~/.claude.json (user scope) or .mcp.json (project scope):

{
  "mcpServers": {
    "gsc": {
      "type": "stdio",
      "command": "node",
      "args": ["/absolute/path/to/google-search-console-mcp/build/index.js"]
    }
  }
}

Claude Desktop

Add to your claude_desktop_config.json:

{
  "mcpServers": {
    "gsc": {
      "command": "node",
      "args": ["/absolute/path/to/google-search-console-mcp/build/index.js"]
    }
  }
}

Usage

Once configured, ask Claude naturally:

• "List my Search Console properties"
• "Show me the top 20 queries for my site over the last 28 days"
• "Check the indexing status of https://example.com/blog/my-post"
• "Compare mobile vs desktop performance this month"
• "What sitemaps are submitted for my site?"

License

MIT