Google Search Console AI Agent

Google Search Console AI Agent

Ask plain-language SEO questions and get back search performance data directly from Google Search Console — no dashboard needed.

What this workflow does

1. SearchQuery — type a free-text SEO question
2. SearchConsoleAgent — an Agent queries GSC through the mcp-search-console MCP server and summarizes the results
3. SearchResult — returns the Agent's answer

Use cases

• "Which pages got the most impressions last week?"
• "What queries rank between position 5 and 15 for my site?"
• "Show me pages with high impressions but low CTR"
• "Is /blog/my-post indexed?"
• "List my submitted sitemaps and their status"
• "What are my top queries by clicks over the last 28 days?"

Service account setup (recommended)

1. Go to Google Cloud Console → IAM & Admin → Service Accounts
2. Create a service account (e.g. [email protected])
3. Enable the Google Search Console API in APIs & Services → Library
4. Download a JSON key for the service account (Actions → Manage keys → Add key → JSON)
5. In Search Console, go to Settings → Users and permissions → Add user, and add the service account email as a Restricted or Full user
6. Base64-encode the key: base64 < service-account.json | tr -d '\n'
7. Paste the result into the GSC_CREDENTIALS_JSON_B64 env field on the MCP connection

OAuth alternative

If you prefer OAuth instead of a service account, set GSC_TOKEN_JSON_B64 (your token.json) and GSC_OAUTH_CLIENT_SECRETS_JSON_B64 (your OAuth client secrets) in the env field instead. The bootstrap script supports both authentication methods.

MCP connection setup

Open SearchConsoleAgent, find the mcp-search-console MCP connection, and update:

• GSC_CREDENTIALS_JSON_B64 — your base64-encoded service account JSON
• Update YOUR_SITE_URL in the system instruction to your verified GSC property URL (e.g. https://example.com/)

Notes

• The Agent is configured read-only: it will not submit sitemaps, delete properties, or make any changes
• The mcp-search-console package is installed automatically by uv run --with
• Requires Python 3.11+ and uv to be installed on the machine running Heym