Why dont we have the parameter of enable thinking or setting the effort in chat model nodes of the ai agent?

Feature Requests
1

I would like to have a special parameter available in all the chat nodes of the AI agent. I think it’s kind of crazy that we don’t have that yet. I should be able to decide whether my AI agent uses reasoning or not, as well as control the reasoning budget and effort level.
I would love to have these parameters available for all providers, like OpenRouter, GPT, Gemini, etc. If you know of a way to configure this, please let me know. I would really appreciate it a lot.

The idea is:

My use case:

I think it would be beneficial to add this because:

Any resources to support this?

Are you willing to work on this?

2

As a workaround until this is added natively: swap the chat model node for an HTTP Request node calling the provider’s API directly. For Anthropic, you can pass "thinking": {"type": "enabled", "budget_tokens": 10000} in the request body. For OpenAI o-series, set "reasoning_effort": "high". Wire it as a tool in your AI Agent by calling a sub-workflow via Execute Workflow node that handles the HTTP call and returns the response. Upvoting this feature request makes sense - direct UI controls for reasoning budget would be much cleaner than the workaround.

3

oi @marcelovaar

the docs shows that some providers/nodes already expose reasoning/thinking parameters, such as OpenAI in Text operations and Google Vertex with Thinking Budget. However, this still does not appear to be standardized across all Chat Model nodes used with AI Agents, nor uniformly available for all providers such as OpenRouter, Gemini, OpenAI, etc.

4

thank u tamy the thing is that im using different models specially chinese ones and im having many issues i dont undenstand why dont n8n just add the http request node for this

5

@marcelovaar ,

With HTTP Request, you can call the provider API directly without depending on which fields are currently exposed in the Chat Model node UI. The main difference is that the Chat Model nodes try to provide a normalized interface across providers, while the HTTP Request node gives full control over the raw API request. So for models that expose custom parameters not yet available in the n8n Chat Model node UI, HTTP Request is often the best workaround.

I think the feature request may be stronger if framed as: “allow advanced/custom provider parameters in Chat Model nodes” or provide a raw JSON/extra parameters field, rather than expecting one universal reasoning setting for all providers.

6

Adding my two cents. It is a bid odd that we have to access Vertex Models as a Chat Model connected to an AI Agent. When you look at the Gemini Node, there are currently 15 actions. I understand that Google has waffled on Gemini standalone versus enterprise GCP/Vertex, but the fact of the matter is that there are all sorts of features that are available when calling models on Vertex that are missing.

I can appreciate calling HTTP request as a workaround, but let’s call it what it is, a workaround.

I am relatively new to this space. Does the dev team discuss roadmap items, or provide a place we where we can upvote feature requests?