A developer has introduced Draco, a new open-source, single-binary web scraping tool written in Rust, designed to serve as a self-hosted alternative to expensive services like Firecrawl or Browserbase. According to Hacker News Front Page, the project aims to address the high memory usage and detection issues associated with traditional headless browser setups. By using a tiered escalation strategy, Draco attempts to bypass bot detection without immediately relying on resource-intensive browser instances.

The tool functions through a tiered architecture: first, it utilizes a 'Stealth Fetch' layer that employs custom TLS and JA4 fingerprinting to mimic human network signatures at the packet level. If a site requires more complex rendering, such as a React-based single-page application, it triggers an in-process V8 engine to extract data from hidden JSON APIs. Only when these methods fail does the software resort to driving a full browser. This approach allows the scraper to operate using significantly less RAM compared to standard frameworks like Playwright.

Beyond basic scraping, Draco includes a daemon mode that provides a REST API compatible with existing industry standards, a built-in Model Context Protocol server for AI integration, and parallel multi-engine web search capabilities. The project is currently available under MIT and Apache-2.0 licenses. While the developer notes that the software is still a work in progress and may face challenges with certain edge-case websites, preliminary benchmarks suggest it maintains high performance on sites protected by security providers like Cloudflare.