mcp-gsc

This MCP server provides an integration between Google Search Console (GSC) and Claude AI, enabling natural language conversations to explore your SEO data. Once running, Claude can help you manage properties, fetch search analytics, inspect URLs, and handle sitemaps by interpreting natural language requests. Available actions include listing properties, getting site details, querying search analytics and performance, checking indexing status, inspecting URLs, and managing sitemaps. You can ask Claude to perform tasks like 'list my properties', 'show me search analytics for example.com last 30 days', or 'submit the sitemap for example.com'. The tools are designed to be used conversationally, but you can also request specific tool calls if needed by explicitly naming the action (e.g., get_search_analytics, check_indexing_issues).

Prerequisites:

• Python 3.11+ and Node.js
• Claude Desktop installed
• Access to Google Search Console (OAuth or Service Account)

1. Clone the repository

git clone https://github.com/AminForou/mcp-gsc.git
cd mcp-gsc

2. Set up Google Search Console credentials

• OAuth (recommended):
  • Create OAuth client in Google Cloud Console and download client_secrets.json
  • Set environment variable GSC_OAUTH_CLIENT_SECRETS_FILE to the path of this file
  • When prompted, authorize access via the browser
• Service Account:
  • Create a service account and download the JSON key as service_account_credentials.json
  • Set environment variable GSC_CREDENTIALS_PATH to the path of this file
  • Add the service account email to the appropriate GSC properties

3. Install required software

• Python (3.11+)
• Node.js
• Claude Desktop

4. Install dependencies (example if server.js uses npm install)

# If there is a package.json for the Node components
npm install

5. Run the MCP server

# Depending on your setup, the server may be started with Node
# Ensure the path in the MCP config points to your actual server file
node path/to/server.js

6. Run the MCP client (Claude) and connect to the running server by configuring your MCP client to the mcp-gsc server as defined in the mcp_config.

Environment variables are placeholders and should point to actual credential files when deploying. If OAuth is used, ensure the client secrets file is accessible and that the redirect/consent screens are properly configured in Google Cloud Console. For service accounts, you must grant access to the relevant GSC properties. If you encounter authentication issues, double-check the file permissions and that the environment variables are correctly exported in the shell that launches the server. The MCP tools support up to 19 distinct functions; you can request any of them by name (e.g., list_properties, get_search_analytics, submit_sitemap).