Calling Browser Agents from n8n workflow

Questions
1

Describe the problem/error/question

I want to call a browser agent (CUA) from an n8n flow. Where it will take some action on a website then return back to me when it is complete or with information. For example, log into site abc.com and check to see what orders are waiting for processing. Clearly this is an example where this not an API or DB access. I also want to be able to download files from one system then upload them to another.

I’m looking at tools like BrowserBase, Skyvern, OpenAI Agent (used to be called operator), Google Mariner.

having a tool like this that can go over the top and access / interact with data would be so powerful.

Anyone tried this or have any tips?

2

I recommend that you use Browser-Use for this. It can take “task” as prompt and then perform the actions you specify. The quality of the task performed is dependent on several factors like LLM used, Prompt and website structure (anti-bots).

3

Will take a look. Curious if others have tried any of these as well. And any pitfalls to avoid.

5

what did you end up choosing?

6

For those of you interested, I ended up building a NodeJS app that bridged to browserbase. It is the right blend of Agentic and Deterministic (via Stagehand/Playwright).

1 Like