Changelog
Product Changelog
Stay up to date with the latest improvements, new features, and bug fixes across the FineData platform.
Latest July 2026
Country targeting everywhere, honest extraction errors
- + Country targeting on the public API: pass
proxy_country(ISO-2, e.g.DE) and it applies to datacenter, residential, ISP and mobile pools alike, on single, async and batch requests. An unsupported country for the resolved pool returns 422 instead of quietly giving you another region. See geo-targeting - ~ AI extraction no longer returns an empty result when the model runs past its budget. You get an explicit reason in
meta.extract_error, so a failed extraction is visible instead of looking like a page with no data - ~ The MCP server exposes the same parameter names as the REST API (
stealth_antibot,use_js_render), so agent code and the docs finally match - ~ Changing or resetting your password now signs out sessions that were issued before the change
- + Documentation: llms.txt for AI agents, comparison pages for the four APIs we get asked about most, and an explicit scope-of-use note on every use-case page
May 2026
Bring Your Own Proxy
- + Save your own proxy list as a profile and pass
proxy_profile_id— you pay a flat +1 token instead of our per-request proxy surcharge, on Team S and above. Details: BYOP - + Per-profile rotation mode, so a BYOP pool can hold one exit IP for a whole browser session the same way our own pools do
April 2026
Page interactions before extraction
- ~
js_actionsnow run in every rendering path, not just the default one — click, type, hover, scroll,wait_for, screenshot andevaluatebehave the same whichever mode the request uses - ~ Fewer false failures: a page that legitimately answers with a short body, or with a 403 that still carries the content, is no longer counted as a blocked render and retried for nothing
March 2026
Proxy and fingerprint controls
- + ISP proxies (
use_isp, +2 tokens): static residential-grade addresses for sites that judge IP reputation - +
proxy_sticky: keep the same exit IP for an entire browser session — required when a warmup request and the request that actually reads the page have to come from one address - + An additional full-browser rendering engine alongside the existing headful path, picked automatically for pages that only render correctly in a complete browser — nothing to change in your request
- ~ TLS impersonation upgraded to curl-cffi 0.15, so
tls_profilekeeps pace with current browser releases;vip:*profiles rotate premium mobile and desktop fingerprints - ~ More residential and ISP capacity behind the same request parameters — no change needed on your side
February 2026
Structured output and AI extraction
- +
formats: ask formarkdown,text,links,rawHtmlor ascreenshotin the same response - + Three ways to get structured data:
extract_rulesfor CSS/XPath,extract_schemafor a JSON Schema, orextract_promptin plain language (+5 tokens) - +
csvandxlsxoutput for AI extraction results, so a table lands as a table - +
only_main_contentstrips navigation, footers, sidebars and ads frommarkdownandtextoutput, so fewer tokens go to boilerplate
February 2026
Platform Launch
- + Initial public release of FineData.ai scraping API
- + Stealth rendering modes for pages that require a full browser
- + JavaScript rendering with headless browser pool
- + Batch and async scraping endpoints
- + MCP server for AI agent integration
Full parameter reference in the API docs. Token costs per feature are on the pricing table.