From MCP to Scale: Pipelines That Build Themselves — Rafael Levi, Bright Data
AI agents using Bright Data's MCP can autonomously write, run, and maintain web scrapers at scale instead of parsing pages with expensive LLM tokens.
“build a scraper that's going to parse it for me”
Bright Data's Rafael Levi demonstrated using AI agents with MCP to build self-maintaining scraping pipelines, where an agent explores a site, writes the scraper, runs it, and auto-fixes breakages rather than using LLMs to parse every page. It matters as a practical pattern for cost-efficient large-scale data collection, but the talk is largely a vendor demo of Bright Data's tooling with limited broad industry significance.