Introducing the /interact endpoint - Just /scrape a page - Call /interact to do any action with natural language - Watch agents click, fill, scroll to get deep web data Available today.

Interact is incredibly easy to use and integrates seamlessly with our scrape and search endpoints. Here's how you can start using it in our CLI.

Every browser session triggered by /interact has a live URL you can embed, share, or interact with in real time. And with persistent profiles, your agent can log in once and pick up wherever it leaves off - no re-authenticating, no lost state.

Need full browser control? Use Playwright code instead. Same page, same session.

Let your agents /interact with the live web:

👀

Let's goooo! Also: Build an agent that uses /interact inside @aisdk (

@firecrawl just killed @comet and @ChatGPTAtlass with an endpoint 😅

There's something addictive about the scrape → interact → scrape loop

interesting concept but how do you handle sites with heavy js/react where the DOM changes dynamically? seen a lot of scrapers break on modern web apps. what's your approach to rendering stability?

Watching agents act on natural language? So wild

that's crazy. you guys should build this as a chrome extension

natural language scraping is finally here lol

sounds powerful. automating interactions could save hours. have you tested its limits or edge cases yet?

I love this.

the death of dom obfuscation - hit the target url - let /interact resolve the auth flow - dump raw session state pure extraction.