unblock

Reading the web with UnblockingAPI

You have the UnblockingAPI MCP server. Use it to read web pages: it goes out through residential proxies and a real browser when needed, so it can read public pages that need JavaScript, refuse simple requests or depend on location.

Workflow

1. Leave render unset. unblock_fetch(url) lets the API decide: a fast plain fetch first, and a real browser instead when that comes back blocked, as a challenge wall or as an empty JavaScript shell. It learns per site, so the next request goes straight to the browser. render_reason in the output says what happened. Clean HTML comes back with scripts, styles, SVGs and cookie banners stripped.
2. Force a mode only when you know better. render: true when the page loaded but content you need is filled in by JavaScript; render: false never renders. A render can take up to two minutes on slow sites.
3. Tune the wait only if a render is still missing content. wait_for takes a CSS selector or stable / domcontentloaded / networkidle; wait_state refines a selector wait (nonempty, hidden for a spinner, stable); scroll_steps triggers lazy sections; wait_min_ms adds a minimum delay; settle_ms caps how long the page may keep changing; wait_rules handles pages with several layouts (consent gate, interstitial). Any of these implies a browser. readiness.verification.status: "interactive" means a verification (Turnstile, reCAPTCHA, DataDome, …; vendor says which) wants a human, and waiting longer will not help.
4. Geo-restricted or localised pages: pass location as a 2-letter country code (us, gb, de, se, …).
5. Re-reading a page: the exact same fetch (another select or format) within 10 minutes is reused for free; pass fresh: true to fetch again. max_age (seconds, up to 300) accepts the API's cached copy instantly but still costs 1 credit, and changes the fetch, so it misses that free reuse.