I'm in the middle of a project where I needed to pull in data from a few dozen help articles. I tried two approaches: using Relevance AI's built-in browser action, and using a separate scraping service (I used Apify) to feed cleaned data into Relevance.
Here's my quick take. The built-in browser is fantastic for quick, ad-hoc tasks. You just give it a URL and it figures out the content. It's zero-config and works right inside your workflow. But for larger jobs, I hit some timeouts and the cost per task added up.
Using a separate scraper gave me more control—I could parse and clean the data exactly how I wanted, cache it, and then send structured JSON to Relevance. It was more initial setup (writing the actor, scheduling runs), but then each AI call was cheaper and more predictable.
For one-off research, I'd stick with the built-in tool. For any automated, recurring data pipeline, I'd now lean towards a dedicated service. Curious if others have hit similar limits or found better ways to tune the browser action.