Skip to main content

I have a list of URLs that return XML documents.

I put the URLs into a new table and created an AI field with the prompt “Fetch the URL at {URL} and process it as follows...” (the rest of the prompt isn’t relevant)

I enabled Internet access in the Airtable Settings and ran the field agent.

Instead of getting the XML documents, I get "I’m unable to fetch or retrieve content from external URLs" (GPT 4.1-mini) or “Sorry, I can't fetch or insert the contents of external URLs.” (GPT 4.1).

I was only able to get this to work if I checked “Enable internet search” for this field, but I do not want it to do a web search. I want it to go fetch a specific URL and process the XML. 

So I have two questions: 

  • Is it actually searching the Web instead of just going directly to the specified URL? (which would be a waste of credits) 
  • If not, why do I need to enable “Web search”?

Thanks for any info,

Natalka

Enabling web search means “let me out of my sandbox and onto the internet”; the tool should just be going to the specific URL if instructed in my experience.


Hey ​@nroshak,

I’m not 100% of how this works on the back. However, in this other post when the launch was announced, ​@chelseytanaka mentions a very similar use case:

  • “Write a summary of the blog post located at this URL: {blog_post_URL}”

Seems like it should be able to do that…

I’ll try to find out more!

Mike, Consultant @ Automatic Nation


Hi ​@nroshak,

 

The AI field when not enabled internet search, all it can do it use the information you provide to it, in this case just that there is a url with the address you provided. When you enable internet search, the AI agent can actually use the internet to get the contents of the url you provided and then analyze that content based on your prompt. 

You can also use a script to fetch the XML and then pass that content to your AI field to process it in the way you want. 
 

I haven’t experimented to see how many tokens it uses to search something on the internet, but potentially you could spend less tokens if using this script approach to fetch the XML and then use the AI field to process it.

 

Hope that helps!

 


Reply