Built for AI agents that need to scrape protected sites without getting blocked. Exposes a unified MCP toolkit covering search, extraction, stateful browser automation, and human-in-the-loop fallback for captchas and auth walls. The browser automation layer replaces Playwright with a single omni-tool that handles navigation, clicks, waits, assertions, network mocking, and screenshot capture in one call, all running in a persistent headless profile. When automated flow hits Cloudflare Turnstile or DataDome, you can invoke the HITL auth portal to solve it once, then the agent resumes silently. Ships as a self-contained Rust binary with validation artifacts for LinkedIn, Ticketmaster, Airbnb, and other heavily protected targets. Reach for this when web scraping needs to survive real-world bot detection at agent speed.
claude mcp add --transport stdio io.github.devshero-shadow-crawl uvx shadow-crawl