Web Search with Anthropic Models?

Questions
1

Hello all,

Following up on the this closed topic: New to n8n - How do I enable web search in Claude Sonnet 4?

From my searching online, and poking around in the nodes, it doesn’t appear to have been added as an option.

I’d really like to be able to send API calls to Anthropic’s models with their web search and thinking enabled, as out of the 10 or so different models I’ve tried across a few different providers, theirs is the only one that has been able to give me the desired information for my particular use case (at least through the web chat, simulating the data that I want to pass to their API).

I’m a brand new user, and I’ve not figured out the HTTP Request Tool node to see if it would fill my needs, but given this is a built-in feature in Anthropic’s API, I believe that it should be added as an option (right alongside the “enable thinking” in the AI model options).

Is this an existing feature that I am missing, or something that can be added to the list of features to add?

Thank you

2

If you use the Anthropic Message a Model node directly, you can set this:

image
image774×884 37.4 KB

So it is simpler than having to make your own HTTP requests.

It is not compatible with the AI agent node though, so you can’t combine with other tools.

To do that one solution is to use the Google Search by Apify as a Tool.

3

Interseting… Good to know, thank you. I just got a response back (a lot more chatty as far as its responses, than ChatGPT or the others I’ve tried - it’s sending back a lot of information about its lookups, along with rationale for decisions. Going to need to figure out how to eliminate or filter that, as it doesn’t seem to be respecting my request to only return JSON). Aside from the Memory feature in the AI Agent (which I don’t need for this use case, anyway), is there any use case where the Chat With A Model falls short vs. the AI Agent?

Would be ideal if they just added that “Web Search” option into the Chat Model node, right alongside the “Enable Thinking,” though, to avoid this confusion for newer users.

1 Like
4

100% with you on that. There’s a Feature Requests topic, might be worth to post it there

Aside from the Memory feature in the AI Agent (which I don’t need for this use case, anyway), is there any use case where the Chat With A Model falls short vs. the AI Agent?

Well the agent can be “Agentic” and choose between any tools you build for it; but when you don’t need to do that only one go of question-answer then the message a model is just as good.