wtf is unblock-web?

unblock-web is a Python library and command-line tool that helps AI agents fetch web pages that would normally fail due to various blocking mechanisms. The problem it addresses: many websites return useless content to automated requests, JavaScript-rendered pages that require a real browser, Cloudflare Turnstile challenges that block bots, ISP-level DNS blocking (the README specifically notes it was field-tested against Indonesian internet censorship, known locally as Internet Positif), and login walls on sites like x.com. Rather than trying one approach blindly, the library uses a four-tier decision tree that picks the right method based on what type of block it detects. Tier 0 is a plain HTTP fetch for simple static pages. Tier 1 uses a stealthy browser called Patchright (a modified Chromium that avoids browser fingerprinting) to handle JavaScript-rendered pages and Cloudflare Turnstile challenges. Tier 2 routes requests through TinyFish, a free geo-proxy service, to bypass ISP-level DNS blocks or geo-restricted content. Tier 3 uses public aggregator mirrors for content that won't render without authentication. In code you call a single fetch function and it automatically selects the appropriate tier, or you can force a specific one. For AI agent integration, the README shows examples of wiring the tool's health check into session startup hooks. The library is installable via pip, available as a Docker container for zero-install use, and requires Python 3.10 or newer. It uses local Chromium rather than paid scraping services and is licensed MIT.