Chat about the Firecrawl Interact APIs.
I’ve been trying to make Firecrawl play agar.io with the interact api but it’s very difficult to have any complex logic be passed through the prompt layer. Would there be a way to decipher user intent for a recurring action on the page and leave the instance running for a bit instead of cutting the scrape session?
LegoGuy, that is awesome! /interact is designed for prompts with a well-defined single, clear task. As long as you reuse the same scrapeId and don’t call stop_interaction, the browser stays open (default: ~10 min total, ~5 min inactivity). Each interact call runs one execution and returns; the session persists for the next call.
For complex or recurring logic like this, use code modenot prompts. Since you’re trying to manage real-time logic, Browser Sandbox is worth a look. This provides a standalone session, configurable TTL, and CDP for Playwright.
The documentation can be really helpful for dissecting issues like this and may even provide more explicit advice for your use case. Highly recommend having your agent take a look (you can use Firecrawl to do this!). Docs: Interact | Firecrawl
1 Like