1. Conduid
  2. Scraping
  3. MCP Google Search Console
MCP server · Scraping

MCP Google Search Console

Google Search Console MCP Pack

Unclaimed blockchain
37Low

Scored 3 days ago · breakdown

About MCP Google Search Console

MCP Google Search Console is an MCP server in the Scraping category: google Search Console MCP Pack.

Install

Claude Code
claude mcp add google-search-console -- npx -y @getmcpads/google-search-console-mcp-server
npx
npx -y @getmcpads/google-search-console-mcp-server
uvx
uvx open-gsc-mcp
pip
pip install open-gsc-mcp

This server has no ConduID identity, so agent calls to it are not receipted. Pin the version you install and review the source before granting it credentials.

Ask AI

✦

Ask AI about MCP Google Search Console

Powered by Claude · Grounded in docs

I know everything about MCP Google Search Console. Ask me about installation, configuration, usage, or troubleshooting.

Security checks

  • ✓README presentThe repository ships documentation.

5 checks have not been run yet.

README

mcp-google-search-console-crunchtools

Secure MCP server for Google Search Console. Query search analytics (clicks, impressions, CTR, position), manage sitemaps, inspect URL indexing status, and manage site properties.

CI PyPI

Installation

uvx (recommended, zero-install)

claude mcp add mcp-google-search-console-crunchtools \
    --env GSC_CLIENT_ID=your_client_id \
    --env GSC_CLIENT_SECRET=your_client_secret \
    --env GSC_REFRESH_TOKEN=your_refresh_token \
    -- uvx mcp-google-search-console-crunchtools

pip

pip install mcp-google-search-console-crunchtools

Container (Podman/Docker)

podman run -d -p 8017:8017 \
    --env-file ~/.config/mcp-env/mcp-google-search-console.env \
    quay.io/crunchtools/mcp-google-search-console \
    --transport streamable-http --host 0.0.0.0

OAuth Setup

This server authenticates to Google using OAuth2 with a refresh token. This is a one-time setup — once you have the three credentials, you store them as environment variables and never need to touch OAuth again.

What you need

Credential What it is Where it comes from
GSC_CLIENT_ID Identifies your OAuth app to Google Google Cloud Console
GSC_CLIENT_SECRET Secret key for your OAuth app Google Cloud Console
GSC_REFRESH_TOKEN Long-lived token that lets the server get access tokens One-time browser consent flow

Step 1: Create a Google Cloud OAuth App

  1. Go to Google Cloud Console
  2. Create a new project (or select an existing one)
  3. Navigate to APIs & Services > Library
  4. Search for Google Search Console API and click Enable
  5. Navigate to APIs & Services > Credentials
  6. Click + CREATE CREDENTIALS > OAuth client ID
  7. If prompted, configure the OAuth consent screen first:
    • User type: External (or Internal if using Google Workspace)
    • App name: anything (e.g., "MCP Search Console")
    • Scopes: add https://www.googleapis.com/auth/webmasters
    • Test users: add your Google account email
  8. Back on Create OAuth client ID:
    • Application type: Desktop app
    • Name: anything (e.g., "MCP Search Console Desktop")
  9. Click Create — copy the Client ID and Client Secret

Step 2: Get a Refresh Token

Run this in your terminal to start the consent flow:

# Set your credentials from Step 1
export GSC_CLIENT_ID="your_client_id_here"
export GSC_CLIENT_SECRET="your_client_secret_here"

# Generate the authorization URL
echo "Open this URL in your browser:"
echo ""
echo "https://accounts.google.com/o/oauth2/v2/auth?client_id=${GSC_CLIENT_ID}&redirect_uri=http://127.0.0.1&response_type=code&scope=https://www.googleapis.com/auth/webmasters&access_type=offline&prompt=consent"
  1. Open the printed URL in your browser
  2. Sign in with the Google account that owns your Search Console properties
  3. Click Allow to grant Search Console access
  4. The browser will redirect to http://127.0.0.1/?code=XXXX&scope=...
  5. This page will fail to load — that's expected. Copy the code= value from the URL bar (everything between code= and &scope)

Now exchange the authorization code for a refresh token:

# Paste the code value from the URL bar (the part between code= and &scope)
AUTH_CODE="paste_your_code_here"

curl -s -X POST https://oauth2.googleapis.com/token \
    -d "client_id=${GSC_CLIENT_ID}" \
    -d "client_secret=${GSC_CLIENT_SECRET}" \
    -d "code=${AUTH_CODE}" \
    -d "grant_type=authorization_code" \
    -d "redirect_uri=http://127.0.0.1" | python3 -m json.tool

The response will include a refresh_token field — copy it. This is the long-lived credential that lets the server authenticate without a browser.

Note: If you don't see refresh_token in the response, add &prompt=consent to the authorization URL and try again. Google only returns the refresh token on the first consent or when explicitly prompted.

Step 3: Store the Credentials

Create an env file:

cat > ~/.config/mcp-env/mcp-google-search-console.env << 'EOF'
GSC_CLIENT_ID=your_client_id
GSC_CLIENT_SECRET=your_client_secret
GSC_REFRESH_TOKEN=your_refresh_token
EOF
chmod 600 ~/.config/mcp-env/mcp-google-search-console.env

How it works at runtime

The server never stores or manages tokens on disk. On each API call:

  1. Server sends the refresh token to Google's token endpoint
  2. Google returns a short-lived access token (valid ~1 hour)
  3. Server uses the access token for the Search Console API call
  4. Access tokens are cached in memory and refreshed automatically when they expire

The refresh token itself never expires unless you explicitly revoke it in your Google Account permissions.

Available Tools (10)

Category Count Tools
Sites 4 list_sites, get_site, add_site, delete_site
Search Analytics 1 query_search_analytics
Sitemaps 4 list_sitemaps, get_sitemap, submit_sitemap, delete_sitemap
URL Inspection 1 inspect_url

Security

  • OAuth2 credentials stored as SecretStr (never logged)
  • Environment-variable-only credential storage
  • Automatic token scrubbing from error messages
  • Pydantic input validation with extra="forbid"
  • No filesystem access, shell execution, or code evaluation
  • TLS certificate validation (httpx default)
  • Request timeouts and response size limits
  • Built on Hummingbird container images

See SECURITY.md for the full security design document.

Development

uv sync --all-extras
uv run ruff check src tests
uv run mypy src
uv run pytest -v
gourmand --full .
podman build -f Containerfile .

License

AGPL-3.0-or-later

README copy captured when this server was indexed; it may lag the repository. The original is authoritative.

Questions

About MCP Google Search Console

How do I install MCP Google Search Console?

Run claude mcp add google-search-console -- npx -y @getmcpads/google-search-console-mcp-server, then add the server to your MCP client's configuration.

Is MCP Google Search Console safe to use with an AI agent?

Its trust score is 37 out of 100 (low). It passes 1 of the 1 static security checks run so far. It has no ConduID identity yet, so agent calls to it are not receipted.

Is MCP Google Search Console still maintained?

Conduid hasn't recorded a commit date for this repository yet. Check the repository directly for recent activity.