Every value /api/explore filters on, each counted within the other filters you send (the products that would match with it): areas, categories, capabilities (with how many an agent did), tags, bot-check vendors (stopped, let in), levels, where the agent stopped, a checkpoint's outcomes, method releases; n is how many match the filters as sent; the first and last drive dates

An area of the category tree (GET /api/explore/options areas), e.g. sales; other: a category in no area yet

A category slug, e.g. crm

A capability id of the shared vocabulary (GET /api/capabilities): products with a task of it; each row then carries capability_state

With capability: only products where our agent's work on it ended so

A derived tag (mcp, openapi, llms-txt, cli, webhooks, x402, mpp, free-plan, no-card-trial, card-trial): products with every tag given

Where our agent stopped, as the product page says (else the first checkpoint it didn't pass on its own); none: it passed every one

With outcome: the state of this checkpoint

With checkpoint: passed (on its own), help (a person did it), stopped, na (nothing to do), unreached (behind a wall), untried

A bot-check vendor our signed agent met (GET /api/explore/options vendors), e.g. cloudflare

With vendor: it stopped our agent, or let it in

Driven on or after

Driven on or before

A method release id (kb/rules/method-releases.json), e.g. 2026-Q4.3

Default: desc for level and date, asc for the rest

Only the open layer is public