BigQuery MCP Server
An MCP server for BigQuery that lets you ask Claude questions about your search and analytics data warehouse and get real answers. Not raw query results. Actual analysis with verdicts and recommendations.
33 tools. GA4 + GSC revenue attribution. Generative AI conversation-query detection. ML forecasting. Anomaly detection. Anonymous query analysis. Free and open source.
Full setup guide with screenshots: suganthan.com/blog/bigquery-mcp-server/
New in v4.1 (AI conversation queries): suganthan.com/blog/ai-mode-queries-search-console/
New in v4.0 (GA4 integration): suganthan.com/blog/google-analytics-bigquery-mcp-server/
See it in action
"How much of my traffic is hidden behind anonymous queries?"
"Forecast my traffic for the next 90 days"
"Are there any traffic anomalies I should investigate?"
"Show me year over year seasonal trends"
"Show me queries where mobile and desktop rank different pages"
"Break down my queries by search intent"
"What are the most common terms in my queries?"
"What new keywords appeared this week?"
What you can ask
What are my quick win keywords?
Which pages are losing traffic and why?
Forecast my traffic for the next 90 days
Are there any traffic anomalies I should investigate?
How much of my traffic comes from anonymous queries?
Find the AI conversation fragments hiding in my queries
Show me year over year seasonal trends
What new keywords appeared this week?
Break down my queries by search intent
Show me queries where multiple pages are competing
Generate a full performance report
How does my CTR compare to benchmarks?
How is my /blog/ cluster performing?
Check for any SEO alerts in the last 7 days
Give me content recommendations
With the GA4 BigQuery export also connected (v4.0+):
Which keywords generate the most revenue?
Which pages have the biggest gap between rankings and conversions?
What is each search position worth on my site?
Which pages have a mismatch between their search snippet and actual content quality?
Compare branded vs non-branded performance for 'my brand, my-brand'
Show me page performance combining search and analytics data
Quick start
git clone https://github.com/Suganthan-Mohanadasan/Suganthans-BigQuery-MCP-Server.git
cd Suganthans-BigQuery-MCP-Server
npm install
npm run build
Add to your Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):
{
"mcpServers": {
"bigquery": {
"command": "node",
"args": ["/path/to/Suganthans-BigQuery-MCP-Server/dist/index.js"],
"env": {
"BIGQUERY_PROJECT_ID": "your-project-id",
"BIGQUERY_KEY_FILE": "/path/to/service-account-key.json",
"BIGQUERY_DEFAULT_DATASET": "searchconsole",
"BIGQUERY_LOCATION": "US"
}
}
}
}
Your service account needs three IAM roles: BigQuery Data Editor, BigQuery Data Viewer, and BigQuery Job User.
Need help with BigQuery setup, bulk export, service accounts, or permissions? The full guide walks through every step with screenshots.
Want the GA4 + GSC blending tools too? Add the GA4 BigQuery export and one extra environment variable (
BIGQUERY_GA4_DATASET). Full walkthrough: GA4 + GSC in BigQuery setup guide.
All 33 tools
BigQuery exclusive (9)
These use BigQuery capabilities the Search Console API simply doesn't have.
GSC analysis (12)
The same analysis tools from the GSC MCP server, rebuilt to run on BigQuery's unsampled data.
GA4 + GSC blending (6, new in v4.0)
Require the GA4 BigQuery export running alongside GSC in the same project. Set BIGQUERY_GA4_DATASET (or pass ga4_dataset per call). Full setup guide.
General purpose (6)
Work with any BigQuery dataset, not just GSC.
Why BigQuery instead of the GSC API?
Both complement each other. Use the GSC MCP server for quick real time lookups. Use this for deeper analysis.
What makes this different
Analysis, not just queries. Most BigQuery tools give you raw SQL access. This ships with 27 pre-built SEO analysis tools: opportunity scoring, cannibalisation detection, decay tracking, CTR benchmarking, traffic drop diagnosis, ML forecasting, AI conversation-query detection, and GA4 + GSC revenue attribution. You ask a question, it runs the analysis and tells you what to do.
ML built in. ARIMA_PLUS traffic forecasting and anomaly detection run directly in BigQuery. No external services, no extra cost beyond standard BigQuery pricing.
Hallucination guardrails. Every tool instructs Claude to base analysis only on returned data. Provenance metadata in every response. Exact numbers, no speculation, insufficient data flagged.
Visual dashboards. Results render as rich, interactive visualisations in Claude Desktop. Summary cards, colour coded indicators, bar charts, and tabbed sections. Not plain text dumps.
Read only. Cost controlled. Only SELECT queries allowed. All mutations blocked. Auto-LIMIT on queries, 10GB bytes billed cap, dry run cost preview. EXPORT, CALL, and EXECUTE statements blocked.
Environment variables
Full guide
Step by step setup with screenshots, cost breakdowns, and honest comparison with dedicated SEO tools:
suganthan.com/blog/bigquery-mcp-server/
Changelog
v4.1.1 Position maths corrected across the 13 gsc_* tools. The bulk export's position columns are zero-based, and the older tools summed them without the + 1 that the ga4_gsc_* tools already applied, so average positions read one rank too good and position-band filters shifted by one (gsc_quick_wins asked for 4 to 15 and effectively selected 3 to 14). All 19 expressions are fixed, lookback windows now anchor to the export's latest data_date rather than the wall clock, and every affected tool description carries a note on how to read avg_position. Independently caught, evidenced and fixed by David Wulf in #1 before this landed; his PR also proposes five new tools, under review for the next release.
v4.1.0 Generative AI conversation queries, the BigQuery twin. gsc_genai_conversation_queries finds the AI conversation fragments hiding in your query data and sorts them into seven kinds: reply artefacts, pivot follow-ups, conversational questions, tracker probes, agent harnesses, pasted strings, and a review pile. Same classifier as the GSC MCP version, run over the bulk export instead of the API, so there are no serving limits on large sites, windows anchor to the export's real freshness, and it reports the anonymised split: how many of your impressions carry no query string at all, which is where most of the conversation pool hides (57.7% on the reference property). John Mueller made the same point on LinkedIn the day it launched: large sites will find more of these queries in the export than any other route returns. Total tools now 33. Full method and findings: the launch post.
v4.0.0 GA4 integration. 6 new tools that blend GA4 conversion and revenue data with GSC search data: ga4_gsc_query_revenue, ga4_gsc_content_roi, ga4_gsc_position_value, ga4_gsc_snippet_mismatch, ga4_gsc_branded_performance, and ga4_gsc_page_performance. Requires the GA4 BigQuery export running alongside GSC and one extra environment variable (BIGQUERY_GA4_DATASET). Total tools now 32. Full setup guide: GA4 + GSC in BigQuery.
v3.1.0 Visual dashboard rendering. All GSC analysis tools now produce rich, interactive visualisations in Claude Desktop with summary cards, colour coded indicators, bar charts, and tabbed sections instead of plain text output. No reinstall needed, just restart Claude Desktop.
v3.0.0 Initial release with 26 tools: 8 BigQuery exclusives (anonymous traffic, ML forecasting, anomaly detection, seasonal analysis, device split, intent breakdown, N-grams, new keywords), 12 GSC analysis tools, and 6 general purpose BigQuery tools.
Licence
Apache 2.0. See LICENSE and NOTICE for details. Use it, fork it, build on it. Just keep the attribution.
Built by Suganthan Mohanadasan. If you find it useful, star it.