Changelog
What changed for Spicrawl users: API behaviour and response shapes, MCP server, CLI and SDK releases. Newest first.
2026-11-16: Requests are never billed above their own price
A challenged request is no longer moved to a dearer browser by itself. Pool requests are retried on another exit, and a request that is still blocked fails at no charge. To let the service try a browser, set max_cost or mode=auto. If you send your own proxy and relied on the automatic browser, set max_cost to 8 or more.
2026-11-16: Block pages are reported as errors, not billed as pages
More challenge and block pages are recognised, including AWS WAF pages, human verification pages and rate limit pages. A page like that returns a challenge error at no charge instead of being billed as a page.
Docs: Errors
2026-11-16: Rendered pages wait for client side content
js_render no longer captures a page while it is still loading its scripts or data. Renders can take 1 to 3 seconds longer, and the content they return is more likely to be complete.
Docs: JavaScript rendering
2026-11-16: Notes in scrape responses
Scrape responses now carry notes: short advice that says when a different option would return more. A plain fetch of a page whose content is built by scripts gets an X-Note header, and a notes entry in the JSON response, that suggests js_render. Failed requests now say why they failed and what to change.
Response headers
Every header Spicrawl returns on /v1/scrape: request id, credits, rendering tier, target status, proxy, cache state, limits and X-Warning codes.
Overview
The four ways an AI agent can use Spicrawl (MCP server, agent skill, CLI, raw HTTP), when to pick each, and how agents read these docs.