by default, it Works with 100+ sites out of the box, but you can also add any site with a search bar at it.

SerpAPI or Firecrawl etc are great for Google results or raw page scraping. SearchResult does something different . . it runs the query on the target site's own search and returns structured results in that site's native schema, auto-detected. You're not getting Google's interpretation of arXiv. You're getting arXiv's actual results, structured and clean. I made wrappers and prompt to detect schema of the result page. and continuously checking for Dom changes for wrapper maintenance. it won't hit llm on run time.

Would love feedback, especially from anyone making Deep research agents.