Prepare your next client review with answers from the BigQuery tables and views your team uses. Catchr MCP lets freelancers and agencies query connected BigQuery data through ChatGPT or Claude, using the columns they select.
Discover the available fields, choose the right date column, and ask for a focused comparison or reporting summary.
“Compare paid media performance across my client accounts and flag the biggest changes.”
Google Ads
GA4
Used by the world's leading companies
BigQuery reporting depends on the schema of each selected table or logical view. Start by discovering its columns, then choose the dimensions, numeric values, and dates that match your client question.
Use discovered dimensions to group the numeric columns you select. Top-level numeric columns are summed, so choose values whose totals make sense for your report.
If a numeric column represents an identifier, explicitly configure it as an identifier before querying. This keeps it available as a text dimension instead of treating it as a value to add up.
Select an available DATE, DATETIME, or TIMESTAMP column to control date filtering. This lets you define which date represents the activity you want to discuss with a client.
Temporal range columns also expose start and end fields. A range start is inclusive and its end is exclusive; filtering one of these endpoints does not find every range that overlaps the reporting period.
Scalar fields inside non-repeated records are available as dotted paths, letting you select nested details without changing the number of source rows.
Repeated fields require more care: select fields along one related array path. Empty or null arrays contribute no rows, and independent arrays cannot be combined in the same query. Numeric values above the selected array level cannot be included as metrics.
A focused BigQuery investigation, a client report, and a question through Catchr MCP serve different needs. Choose the workflow that fits the answer or deliverable you need.
Use a report for a repeatable client presentation and Catchr MCP for questions about the connected BigQuery data behind it. Keep the selected columns, date field, and aggregation consistent so the answers are comparable.
Authorize the BigQuery source through OAuth2 and select one data project, then choose the tables or logical views you want to query. Source authorization is separate from the service-account setup used for BigQuery destination exports.
Discover the available columns, select a date field when you need date filtering, and configure any numeric identifiers. Connect Catchr MCP to ChatGPT or Claude to start asking questions. The billing project defaults to the selected data project; connection options include a billing-project override and a maximum bytes billed setting.
Adapt these prompts to a connected client table or view. Each starts with field discovery so the analysis uses columns that actually exist in your BigQuery schema.
Using Catchr, discover the columns in [table or view] for [client connection]. Ask me to select a reporting dimension, a numeric column suitable for summing, and an available date column. For [start date] to [end date], summarize the selected numeric column by the selected dimension and draft a short client-call brief. State the columns, date filter, and aggregation used.
Copy promptUsing Catchr, discover the fields in [table or view] for [client connection], then ask me to choose a dimension, an additive numeric column, and a date column. Compare totals by dimension for [first period] and [second period]. Flag absolute changes greater than [threshold], show both period totals, and describe the differences without assuming their causes.
Copy promptUsing Catchr, discover the columns and their reporting roles in [table or view] for [client connection]. Ask me to select an identifier, an additive numeric column, and a date column. If the identifier is numeric, confirm that it is configured as a text identifier before querying. For [start date] to [end date], show totals by the selected identifier and flag totals above [threshold].
Copy promptUsing Catchr, discover the repeated fields in [table or view] for [client connection]. Ask me to select one supported array path, a dimension and a numeric field within that path, and an available date field for [start date] to [end date]. Summarize the numeric field by the selected dimension. Explain that empty or null arrays contribute no rows, and stop to clarify if the selection combines independent arrays or a metric above the selected array level.
Copy promptThe BigQuery source supports tables and logical views in one configured data project. Available fields come from the selected schema, so discover the columns before requesting an analysis. Custom SQL is excluded.
Authorize the source with OAuth2, select the data project and table or view, then choose the fields for reporting. Select an explicit date column for date filtering. An existing BigQuery destination service-account configuration does not replace source authorization.
Top-level numeric columns are treated as metrics and summed. Explicitly configure numeric identifiers as identifiers so they are treated as text dimensions. Choose additive values for totals; do not assume every numeric column is meaningful when summed.
Multi-account field choices are limited to shared column names with compatible types and reporting meanings. Date choices also require matching names and types. Discover the common fields before comparing data; selecting multiple tables does not establish a join between their records.
Connect your sources to Catchr, add the MCP server to your AI assistant, and ask about the work that matters now.
