How to Connect BigQuery to Claude (1 Minute, No Code)

To connect BigQuery to Claude, link your project at onboard.windsor.ai/app/big_query, then add the Windsor.ai connector in Claude. That’s it, open a new chat and ask questions about your data in plain English.

Here’s the bottleneck it removes: BigQuery already holds some of the most accurate answers in the company, but getting one out usually means writing SQL, waiting on an analyst, or building a BI dashboard just to settle a single question.

Marketing wants to know which campaign produced the highest-value customers last quarter. Product wants to know which feature correlates with 90-day retention. Finance wants a margin breakdown by region for the current month. Each of those is a query away, but “a query away” usually means a ticket in someone else’s queue.

With Windsor MCP for Claude, that queue disappears. Windsor connects directly to your BigQuery project and makes your datasets, tables, and views instantly available in Claude, in less than a minute, with no code.

2 steps to connect BigQuery to Claude

Windsor MCP acts as the bridge between your BigQuery project and Claude. It reads your actual dataset and table structure, not a fixed field list, so Claude works with the same schema your analysts already use. Setup takes two simple steps.

📖 Full documentation: How to integrate data into Claude with Windsor MCP.

Prerequisites

• A Windsor.ai account (free or paid plan)
• A BigQuery project with a service account (or a Google account with OAuth access) that has read access to the datasets you want to expose
• A Claude account

Step 1. Connect BigQuery to Windsor

Go to onboard.windsor.ai/app/big_query and select BigQuery as your data source. Authenticate with a service account JSON key or your Google account, granting read-only access to the project. Then choose the datasets and tables you want Windsor to make available. Windsor fetches the schema automatically, no manual mapping required.

Step 2. Add the Windsor.ai connector in Claude

In Claude, open the Windsor.ai connector page and click Connect.

💡 Quick tip: Set permissions to “Always allow” so your BigQuery tables are live in every Claude conversation automatically. You can also ask Claude to work with only the tables relevant to a given analysis, instead of exposing your entire project.

To confirm the connection works, start with a simple test prompt:

List the tables available in my BigQuery connection and show me a sample of 5 rows from the first one.

If Claude returns real table names and real rows from your project, the connection is live and ready for deeper analysis.

What you can do with BigQuery data in Claude

BigQuery isn’t a fixed-field connector like an ad platform or CRM, it’s your own data warehouse, and every project is structured differently. Windsor doesn’t map your tables to a predefined list of metrics; instead, it reads your actual project structure: datasets, tables, columns, and views, exactly as they exist in BigQuery.

That means Claude can see whatever schema you’ve built, whether that’s a set of dbt models named fct_orders and dim_customers, or raw event tables streamed from GA4 or an app backend. Add a new table or column to your project, and it’s available in the next Claude conversation with no reconfiguration on your end.

What to ask Claude about your BigQuery data: prompt ideas

The prompts below are generic on purpose, since BigQuery has no fixed schema. Swap in your own table and column names, and the same analytical logic applies.

🤖 For more prompt ideas across various data sources, browse the Windsor AI Prompt Library.

List all tables and their row counts

List all tables in this BigQuery project along with their row counts and last modified date.

Based on table and column names, which tables look most relevant for revenue or customer analysis?

Analyze month-over-month trend in a core metric

Using the [orders] table, calculate total revenue and order count by month for the last 12 months.

What is the month-over-month growth rate?

Flag any month where revenue dropped by more than 10% compared to the prior month, and suggest a likely cause based on order volume vs. average order value.

Join two tables to answer a cross-table question

Join the [customers] table with the [orders] table on customer_id.

For each acquisition_channel, show total customers, total revenue, and average order value.

Which channel has the highest average order value? Which has the most customers but the lowest average order value?

Audit data quality across tables

For the [users], [orders], and [events] tables, check for:
- NULL values in key columns
- Duplicate primary keys
- Date columns with values outside a plausible range (e.g. future dates)

Summarize the top data quality issues by potential impact on analysis, and suggest which one to fix first.

Bonus: analyze BigQuery alongside your full marketing stack

Windsor connects to 350+ data sources, so your BigQuery tables don’t have to sit in isolation in Claude. Combine them with live data from the rest of your marketing stack in the same conversation:

• BigQuery + GA4: join warehouse tables with live GA4 behavioral data in one query.
• BigQuery + Google Ads: blend warehouse revenue data with live ad spend.
• BigQuery + Google Sheets: bring warehouse query results into a shareable sheet-style view.

Conclusion

BigQuery already holds a clear picture of the business. The problem has always been the number of steps between a question and an answer: write the query, wait for an analyst, or build a dashboard just to check one number.

Windsor MCP removes those steps. Claude gets live access to your actual BigQuery project, your real datasets and tables, not a simplified summary of them. The analysis that used to take a SQL query and a wait now takes a single sentence.

Whether you’re checking one number or running a full cross-table analysis, the warehouse is now as easy to talk to as a colleague.

🚀 Connect your BigQuery project to Claude with Windsor MCP in less than a minute. Start for free now: https://onboard.windsor.ai/app/big_query.