Latest updates and improvements to FourA
Activity now fits a 14" MacBook without sideways scrolling, and every column keeps its value. The detail panel is a tabbed pane (Request, Response headers, Response body), so the answer is at the top instead of 900px down. Tooltips follow your pointer and read out the full URL.
The method column now reads the verb from your request body (GET, POST, PUT, etc.), instead of always saying POST because that's what our API layer took. Playground calls also record the browser IP of the person who clicked Send, so a shared account can tell who ran what.
A saturated instance answers a health check in 3 to 10 seconds where a healthy one takes 9 to 225 milliseconds, and both used to read as fine. Slow now reads as degraded, carrying the instance's own reason, so we see the problem while it's still one slow server and not yet your failed request. Auto is watched too, with limits set to how long a real hard solve takes instead of a generic one that would fire on every success.
The Dashboard Playground now has three cascading selects for OS, browser, and version, pulled live from the profile catalogue. Pick any two and the third narrows to whatever actually resolves. Choosing a profile turns unblocker on, because the API refuses a profile without it.
The default signature the unblocker sends is now a current major, not the old one scrapers hardcoded. Sites that had started returning static block pages to older majors return their real page on Single and Browser. Both products moved together, so a cf_clearance replay from Browser to Single still lands byte-for-byte.
Browser used to charge the 30-credit defense-solved tier only when a Cloudflare clearance cookie showed up. Now it does the same for any defense we recognise and clear, starting with SiteGround. The response also carries defenses: { present, cleared }, so you can tell which vendor was in the way even when we haven't beaten it yet (DataDome, PerimeterX, Akamai, Incapsula and hCaptcha are recognised for now, not yet cleared).
Browser responses now return defenses: { present, cleared }: every vendor the target ran, and every one we actually cleared. Billing follows the same rule. A request that clears any vendor (not only Cloudflare) is a defense solve, and vendors we recognise but can't yet beat show up in present and never raise the price.
If you pass exitCountries and your plan doesn't include geo targeting, you'll get a 403 that names the parameter and points to the fix. No more guessing which field on the request tripped the check. Plans that do include geo targeting are unaffected.
The plan compare table now has a bandwidth row. Every public plan comes with unlimited bandwidth, so egress won't quietly eat your budget as your traffic scales. Custom plans show their exact GB figure when they're capped.
Any worker that dropped out between releases used to stay offline until someone noticed, quietly dragging peak-hour capacity. Now every rollout brings the full pool back to strength alongside the rolling update, so traffic peaks land on the capacity you expect.
Usage & Limits now has per-tab attention dots, colored usage bars (amber at 80%, red once you're over), a sortable Last used column and 30-second auto-refresh. Open it from the sidebar in your Dashboard.
The pricing page compare table is now grouped by usage, features, rate limits, and account. Every cell comes from the same plan model the API reads, so what you see there is what your plan actually includes. Bandwidth is uncapped on every public plan.
Proxy Finder now runs its three liveness checks concurrently instead of one after another, so dead exits get recycled roughly three times faster. A head-of-line fix keeps validation throughput steady when one subnet hits its rate limit. Expect fewer stale proxies picked for your requests.
Proxies now get validated continuously from a separate network path, so the working pool only holds exits that actually work right now (not just when they were first found). When you call /api/proxy or let /api/auto route your request, the picked proxy is more likely to work first try.
The API blocks any Single or Browser request whose target URL, follow-redirect chain, or proxy field resolves to a private or reserved IP range. Even if a public target redirects into an internal network, the request stops before it gets there.
