One rule across every actor: a row is written and billed only when it contains data.

Almost nobody else does the next part: each skipped page comes back in a free list with a reason code — ROBOTS_DISALLOW, HTTP_403, NO_STRUCTURED_DATA, NETWORK_ERROR. You learn why a record is missing, not just that it is.

• No start fee, no minimum charge, no platform usage line. The actor page price is the whole invoice.
• Failures are loud; robots.txt is fail-closed. A counter mismatch, a parse error or zero records marks the run FAILED. If robots.txt cannot be read, nothing is fetched.
• Measured, not claimed. 150 SimilarWeb domains with 30 fakes planted: 115 written, 35 reported free, 0 parse errors. 400 Hepsiburada pages: 400 HTTP 200s, 13,734 products, 0 blocks. 600 Gumroad products: 600 unique, each billed once.

Newest: Tech Stack & DNS Intelligence. It reads a domain’s DNS as well as its HTML, so it finds the back-office software the page never shows — 7.5 products per domain, measured — and still returns a record when the site answers 403.

	#	Price	Sources
Commerce	2	$0.50–$2.00/1k	Hepsiburada · Gumroad
Web intelligence	2	$6.00/1k · $0.05 per domain	SimilarWeb · Tech Stack + DNS
Jobs and events	2	$1.50–$2.50/1k	Eventbrite · Working Nomads
Education	1	$2.50/1k	Coursera

Each actor's own page carries its field list, its measured run and its price. 7 live.

No names, no contact details, no identifiers. Output comes from a published field whitelist, and where a source attaches a person id to a public review, that id is dropped before the record is written. Every actor runs as an AI agent tool over the Apify MCP server.

Built by a QA engineer, five years in test automation. A scraper that quietly returns nothing is a test that passes without asserting.