Don't know if it's called refunds or credit_memos. Describe what the table should contain and let the Ask panel find the real name.
No credit card. 14 days. Cancel in one click.
Quick answer: Use the explorer's Filter box when you know roughly what a table or column is called. Use the Ask panel when you know what the data represents but not the name, ask in plain English and it scans table names, column names, and comments to find the match. Right-click the result for a starting query or Insert table name.
A schema with a few dozen tables is easy to scan by eye. A schema with a few hundred, spread across datasets or catalogs with inconsistent naming from years of different teams, isn't. QueryFlow gives you two ways in: a direct filter for when you know roughly what you're looking for, and a plain-English search through the Ask panel for when you know what the data means but not what it's called.
Say you inherited a Snowflake warehouse and need "whatever table tracks refunds," but you don't know if it's called refunds, credit_memos, or something from an old system entirely. Ask the panel: "which table tracks customer refunds." It scans table and column names, and comments where available, and comes back with the actual table, something like finance.credit_memos, along with a short note on why it matched.
Right-click the table in the explorer for a starting query, SELECT * FROM finance.credit_memos LIMIT 100, or choose Insert table name to drop the reference straight into your editor at the cursor. Copy name is useful when you're writing a query by hand and just need the exact identifier, case included.
If the Filter box or the Ask panel surfaces the right table on the first try, and clicking into it shows columns that match what you expected, the search did its job. If nothing matches, the table may live in a schema or dataset you haven't been given access to yet, worth checking with whoever manages the connection.
| If you see | Fix |
|---|---|
| The explorer shows no schema at all | Click Retry, or Test the connection first, per connection diagnostics. |
| Filter finds nothing for an obvious name | Check for a typo, or that the table lives in the schema/dataset currently expanded, not a sibling one. |
| Ask returns the wrong table among several similar ones | Add a distinguishing detail, like an approximate column name or the team that owns it. |
| A very large schema loads slowly | This is normal on catalogs with thousands of tables. The filter narrows results as you type without needing the whole tree expanded. |
It searches structure, table names, column names, comments, not the data inside the tables. Finding which rows contain a value is a WHERE clause, not a schema search.
On a BigQuery project with datasets from several teams, ask "which dataset has anything related to marketing attribution." The panel checks dataset and table names across what your account can see and points you toward the likely one, saving the click-through-everything approach on a project you didn't set up yourself.
A five-table schema doesn't need this: you remember what's in it. A schema that's grown for years across several teams, with tables named by whoever set them up at the time, is where structural search stops being a convenience and starts being the only practical way in for someone new to the warehouse.
Filter matches names directly, fast when you know roughly what the table or column is called. Asking Claude works when you know what the data represents but not what it's actually named.
It reads whatever metadata your database exposes, including comments where the source supports them, in addition to names.
No, schema search is scoped to the connection you're currently browsing, one warehouse or database at a time.
It's most useful once a schema has enough tables that you can't hold them all in memory, but it works fine on a small one too.
The Filter box doesn't. Asking Claude to find something in plain English does, since that's the Ask panel's schema-grounded search.
14-day free trial, no card. Try it on your own warehouse.
No credit card. 14 days. Cancel in one click.