Credit costs
Each data extraction endpoint consumes a fixed number of credits per successful request:Credits over MCP
The MCP server charges exactly what the equivalent REST endpoint charges —prefetch_enrich costs 20 credits, prefetch_scrape costs 3, and so on. There is no surcharge for calling a tool instead of an endpoint.
Two MCP tools are free, because they only read work you have already paid for:
prefetch_check_crawl_status— bills only the crawl pages that completed since your last call.prefetch_cancel_crawl— never charges.
When credits are deducted
Credits are deducted only on successful responses — meaning the API returned"success": true with a 200 status code.
You are not charged for:
- Failed requests (4xx or 5xx responses)
- Validation errors (missing URL, bad parameters)
- Authentication failures
- Rate limit rejections (429)
- Timeouts (504)
- Individual crawl pages that failed to fetch (
ok: false)
How crawls are charged
A crawl is the only endpoint billed across more than one request:- 3 credits when you submit it with
POST /crawl. - 3 credits per page the crawl actually returns, charged on
GET /crawl/{id}as pages complete.
- You pay for pages fetched, never for the
limityou requested. A crawl capped at 500 pages that finds 12 costs 36 credits, not 1,500. - Polling is free. Each status call bills only the pages that completed since your previous call, so checking every five seconds costs exactly the same as checking once at the end.
Credit limits
Each API key can have acredits_limit configured. When cumulative credits_used reaches this limit, all further requests return:
Choosing the right endpoint
When you need brand, company, and classification data for a URL, use/enrich for a single optimised call instead of three separate requests:
Use /enrich
20 credits for brand + company + classification.Best when you want a complete domain profile in one call.
Use individual endpoints
More targeted — only pay for what you need.Best when you only need one or two data types.