Skip to main content
You do not need to know the catalog. Describe what you want and discover ranks the tools against it.

Options

Plus the global options.

Reading the price column

The PRICE column is the rate: what one unit costs you, printed flat because it is exact. A row, a page, a thousand bytes of text, a second of audio, or the whole call on a flat-priced image tool. Two of those are not the total. On a scrape the total follows the rows the call really produced, and on a voice call it follows the size of what you sent. The words “up to” appear for a ceiling and for nothing else, so a number without them is never a maximum.
A rate is not a total. A web.search call at 0.0075perpagechargesaroundacentforanormalrequest,wellunderthe0.0075 per page charges around a cent for a normal request, well under the 1.00 ceiling inspect shows. An image call is the other extreme: one exact price, nothing to settle. See Pricing.

JSON output

per_unit_usd and unit_label are the rate the table prints. billed_on_actual_usage: true says the total still depends on the call, so max_usd next to it is a maximum. On a flat-priced tool that flag is absent and base_usd is the whole price. An agent can tell the two apart without reading English. sample_input is fillable, so you can go straight to run without an inspect round trip.

No matches

Exit code is still 0. An empty catalog match is an answer, not an error.

Next: inspect

Read the exact input a tool takes before you call it.