Clarification on Tool Behavior: Is `curl` in AI Studio Using a Cached Web Version?

Google AI Studio
1

Hi,

I’m prototyping an agentic workflow in a AI Studio and have a question about the behavior of the curl tool.

I’ve observed a discrepancy:

1. When I use curl"example-expired-domain.comin the sandbox, it successfully returns HTML content, seemingly from a time when the site was live.

2. When I try to access the same URL in my browser or from my local terminal, I get a “Could not resolve host” error, as expected for an expired domain.

This leads me to believe that the curl tool within the sandbox is not making a live request to the internet but is instead interacting with a cached or pre-indexed version of the web.

Could you please confirm if this is the intended behavior? Understanding whether the sandbox tools are interacting with the live internet or a cached version is critical for developing reliable agents.

Thank you

3

Hello,

To better assist you, could you please clarify what you mean by the ‘curl’ tool?

If you are attempting to have the model access content from a URL, you can use the URL context tool available in the Run settings panel in Google AI Studio. When this tool is enabled, the model can read and extract information from the URL provided in your prompt.

However, if the tool is disabled, the model might generate a response from its own knowledge without accessing the URL. This could be why it is providing a response for a URL that does not exist.

If this does not resolve the issue, please provide the exact prompt and model configurations from your AI Studio chat session so we can better assess the problem.